Siga este link para ver outros tipos de publicações sobre o tema: Potentiel machine learning.
Teses / dissertações sobre o tema "Potentiel machine learning"
Crie uma referência precisa em APA, MLA, Chicago, Harvard, e outros estilos
Veja os 50 melhores trabalhos (teses / dissertações) para estudos sobre o assunto "Potentiel machine learning".
Ao lado de cada fonte na lista de referências, há um botão "Adicionar à bibliografia". Clique e geraremos automaticamente a citação bibliográfica do trabalho escolhido no estilo de citação de que você precisa: APA, MLA, Harvard, Chicago, Vancouver, etc.
Você também pode baixar o texto completo da publicação científica em formato .pdf e ler o resumo do trabalho online se estiver presente nos metadados.
Veja as teses / dissertações das mais diversas áreas científicas e compile uma bibliografia correta.
1
Artusi, Xavier. "Interface cerveau machine avec adaptation automatique à l'utilisateur". Phd thesis, Ecole centrale de Nantes, 2012. http://www.theses.fr/2012ECDN0018.
Texto completo da fonteResumo:
Nous nous intéressons ici à une interface cerveau-machine (BCI, Brain Computer Interface) permettant de commander une prothèse par la pensée. Le rôle du BCI est de décoder à partir de signaux électroencéphalographiques (EEG) le mouvement désiré par le sujet. Le cœur du BCI est un algorithme de classification caractérisé par le choix des descripteurs des signaux et des règles de décision. L’objet de cette thèse est de développer un système BCI précis, capable d’améliorer ses performances en cours d’utilisation et de s’adapter à l’utilisateur sans nécessiter de multiples sessions d’apprentissage. Nous combinons deux moyens pour y parvenir. Le premier consiste à augmenter la précision du système de décision en recherchant des descripteurs pertinents vis à vis de l’objectif de classification. Le second est d’inclure un retour de l’utilisateur sur le système de décision : l’idée est d’estimer l’erreur du BCI à partir de potentiels cérébraux évoqués, reflétant l’état émotionnel du patient corrélé au succès ou à l’échec de la décision prise par le BCI, et de corriger le système de décision du BCI en conséquence. Les principales contributions de la thèse sont les suivantes : nous avons proposé une méthode d’optimisation de descripteurs à bases d’ondelettes pour des signaux EEG multivoies ; nous avons quantifié théoriquement l’amélioration des performances apportée par le détecteur ; un simulateur du système corrigé et bouclé a été développé pour observer le comportement du système global et comparer différentes stratégies de mise à jour de l’ensemble d’apprentissage ; le système complet a été implémenté et fonctionne en ligne dans des conditions réellesWe study a brain computer interface (BCI) to control a prosthesis with thought. The aim of the BCI is to decode the movement desired by the subject from electroencephalographic (EEG) signals. The core of the BCI is a classification algorithm characterized by the choice of signals descriptors and decision rules. The purpose of this thesis is to develop an accurate BCI system, able to improve its performance during its use and to adapt to the user evolutions without requiring multiple learning sessions. We combine two ways to achieve this. The first one is to increase the precision of the decision system by looking for relevant descriptors for the classification. The second one is to include a feedback to the user on the system decision : the idea is to estimate the error of the BCI from evoked brain poten tials, reflecting the emotional state of the patient correlated to the success or failure of the decision taken by the BCI, and to correct the decision system of the BCI accordingly. The main contributions are : we have proposed a method to optimize the feature space based on wavelets for multi-channel EEG signals ; we quantified theoretically the performances of the complete system improved by the detector ; a simulator of the corrected and looped system has been developed to observe the behavior of the overall system and to compare different strategies to update the learning set ; the complete system has been implemented and works online in real conditions
2
Artusi, Xavier. "Interface Cerveau Machine avec adaptation automatique à l'utilisateur". Phd thesis, Ecole centrale de nantes - ECN, 2012. http://tel.archives-ouvertes.fr/tel-00822833.
Texto completo da fonteResumo:
Nous nous intéressons ici à une interface cerveau-machine (BCI, Brain Computer Interface) permettant de commander une prothèse par la pensée. Le rôle du BCI est de décoder à partir de signaux électroencéphalographiques (EEG) le mouvement désiré par le sujet. Le coeur du BCI est un algorithme de classification caractérisé par le choix des descripteurs des signaux et des règles de décision. L'objet de cette thèse est de développer un système BCI précis, capable d'améliorer ses performances en cours d'utilisation et de s'adapter à l'utilisateur sans nécessiter de multiples sessions d'apprentissage. Nous combinons deux moyens pour y parvenir. Le premier consiste à augmenter la précision du système de décision en recherchant des descripteurs pertinents vis à vis de l'objectif de classification. Le second est d'inclure un retour de l'utilisateur sur le système de décision : l'idée est d'estimer l'erreur du BCI à partir de potentiels cérébraux évoqués, reflétant l'état émotionnel du patient corrélé au succès ou à l'échec de la décision prise par le BCI, et de corriger le système de décision du BCI en conséquence. Les principales contributions de la thèse sont les suivantes : nous avons proposé une méthode d'optimisation de descripteurs à bases d'ondelettes pour des signaux EEG multivoies ; nous avons quantifié théoriquement l'amélioration des performances apportée par le détecteur ; un simulateur du système corrigé et bouclé a été développé pour observer le comportement du système global et comparer différentes stratégies de mise à jour de l'ensemble d'apprentissage ; le système complet a été implémenté et fonctionne en ligne dans des conditions réelles.
3
Ohlsson, Caroline. "Exploring the potential of machine learning : How machine learning can support financial risk management". Thesis, Uppsala universitet, Företagsekonomiska institutionen, 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-324684.
Texto completo da fonteResumo:
For decades, there have been developments of computer software to support human decision making. Along with the increased complexity of business environments, smart technologies are becoming popular and useful for decision support based on huge amount of information and advanced analysis. The aim of this study was to explore the potential of using machine learning for financial risk management in debt collection, with a purpose of providing a clear description of what possibilities and difficulties there are. The exploration was done from a business perspective in order to complement previous research using a computer science approach which centralizes on the development and testing of algorithms. By conducting a case study at Tieto, who provides a market leading debt collection system, data was collected about the process and the findings were analyzed based on machine learning theories. The results showed that machine learning has the potential to improve the predictions for risk assessment through advanced pattern recognition and adapting to changes in the environment. Furthermore, it also has the potential to provide the decision maker with customized suggestions for suitable risk mitigation strategies based on experiences and evaluations of previous strategic decisions. However, the issues related to data availability were concluded as potential difficulties due to the limitations of accessing more data from authorities through an automated process. Moreover, the potential is highly dependent on future laws and regulations for data management which will affect the difficulty of data availability further.
4
Hu, Jinli. "Potential based prediction markets : a machine learning perspective". Thesis, University of Edinburgh, 2017. http://hdl.handle.net/1842/29000.
Texto completo da fonteResumo:
A prediction market is a special type of market which offers trades for securities associated with future states that are observable at a certain time in the future. Recently, prediction markets have shown the promise of being an abstract framework for designing distributed, scalable and self-incentivized machine learning systems which could then apply to large scale problems. However, existing designs of prediction markets are far from achieving such machine learning goal, due to (1) the limited belief modelling power and also (2) an inadequate understanding of the market dynamics. This work is thus motivated by improving and extending current prediction market design in both aspects. This research is focused on potential based prediction markets, that is, prediction markets that are administered by potential (or cost function) based market makers (PMM). To improve the market’s modelling power, we first propose the partially-observable potential based market maker (PoPMM), which generalizes the standard PMM such that it allows securities to be defined and evaluated on future states that are only partially-observable, while also maintaining the key properties of the standard PMM. Next, we complete and extend the theory of generalized exponential families (GEFs), and use GEFs to free the belief models encoded in the PMM/PoPMM from always being in exponential families. To have a better understanding of the market dynamics and its link to model learning, we discuss the market equilibrium and convergence in two main settings: convergence driven by traders, and convergence driven by the market maker. In the former case, we show that a market-wise objective will emerge from the traders’ personal objectives and will be optimized through traders’ selfish behaviours in trading. We then draw intimate links between the convergence result to popular algorithms in convex optimization and machine learning. In the latter case, we augment the PMM with an extra belief model and a bid-ask spread, and model the market dynamics as an optimal control problem. This convergence result requires no specific models on traders, and is suitable for understanding the markets involving less controllable traders.
5
Gustafson, Jonas. "Using Machine Learning to Identify Potential Problem Gamblers". Thesis, Umeå universitet, Institutionen för tillämpad fysik och elektronik, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-163640.
Texto completo da fonteResumo:
In modern casinos, personnel exist to advise, or in some cases, order individuals to stop gambling if they are found to be gambling in a destructive way, but what about online gamblers? This thesis evaluated the possibility of using machine learning as a supplement for personnel in real casinos when gambling online. This was done through supervised learning or more specifically, a decision tree algorithm called CART. Studies showed that the majority of problem gamblers would find it helpful to have their behavioral patterns collected to be able to identify their risk of becoming a problem gambler before their problem started. The collected behavioral features were time spent gambling, the rate of won and lost money and the number of deposits made, all these during a specific period of time. An API was implemented for casino platforms to connect to and give collected data about their users, and to receive responses to notify users about their situation. Unfortunately, there were no platforms available to test this on players gambling live. Therefore a web based survey was implemented to test if the API would work as expected. More studies could be conducted in this area, finding more features to convert for computers to understand and implement into the learning algorithm.
6
Del, Fré Samuel. "Études théoriques de la photodésorption d'analogues de glaces moléculaires interstellaires : application au monoxyde de carbone". Electronic Thesis or Diss., Université de Lille (2022-....), 2024. http://www.theses.fr/2024ULILR039.
Texto completo da fonteResumo:
Des quantités inhabituelles de molécules en phase gazeuse sont détectées dans les régions froides (environ 10 K) du milieu interstellaire (ISM), principalement attribuées à la désorption non thermique de molécules depuis les glaces déposées sur les grains de poussière. En particulier, la désorption induite par les rayons ultraviolets du vide (photodésorption VUV) est considérée comme étant une voie de désorption majoritaire dans les régions de l'ISM dominées par les photons. Les investigations expérimentales ont révélé que dans les glaces pures de monoxyde de carbone (CO), espèce omniprésente dans l'ISM, la photodésorption VUV peut suivre un mécanisme indirect de désorption induite par transition électronique (DIET) pour les photons dont l'énergie est comprise entre 7 et 10 eV. Néanmoins, la compréhension des mécanismes moléculaires sous-jacents reste un sujet de débat scientifique. Dans ce contexte astrochimique, nous présentons une étude théorique combinée utilisant la dynamique moléculaire ab initio (AIMD) basée sur la théorie de la fonctionnelle de la densité (DFT) et des potentiels machine learning (PML) construits avec des réseaux de neurones artificiels (ANN), afin d'étudier la dernière partie du mécanisme DIET dans les glaces amorphes de CO. Ici, une molécule CO hautement excitée vibrationnellement (v = 40) au centre d'un agrégat composé de 50 molécules de CO, initialement optimisé puis thermalisé à 15 K, déclenche, la désorption indirecte de molécules de surface. Nos résultats théoriques révèlent que le processus de désorption consiste en 3 étapes fondamentales qui commence par une attraction mutuelle entre la molécule excitée vibrationnellement et une ou deux molécules voisines, activée par l'étirement de la liaison CO et favorisée par l'effet stérique des molécules environnantes. Cela est suivi par une séquence de transferts d'énergie initiée par une collision, se concluant en la désorption de molécules CO vibrationnellement froides dans 88% des trajectoires AIMD. De plus, les distributions théoriques de l'énergie interne et translationnelle des molécules désorbées concordent remarquablement avec les résultats expérimentaux, ce qui soutient le rôle crucial de la relaxation vibrationnelle dans le processus de désorption. Enfin, les premiers PML construits à partir des simulations AIMD, sont capables d'ajuster avec précision la surface d'énergie potentielle multidimensionnelle du système, permettant de prédire efficacement les énergies des agrégats et les forces atomiques. Les simulations de dynamique moléculaire classique utilisant ces potentiels sont plus de 1800 fois plus rapides que celles basées sur l'AIMD, tout en offrant des précisions similaires à ceux de la DFTUnusual amounts of gas-phase molecules are detected in the cold regions (around 10 K) of the interstellar medium (ISM), primarily attributed to the non-thermal desorption of molecules from ices deposited on dust grains. In particular, vacuum ultraviolet (VUV) photon-induced desorption (photodesorption) is considered a major desorption pathway in photon-dominated regions of the ISM. Experimental investigations have revealed that in pure carbon monoxide (CO) ices, a ubiquitous species in the ISM, VUV photodesorption can follow an indirect mechanism of desorption induced by electronic transitions (DIET) for photons with energy between 7 and 10 eV. Nevertheless, the understanding of the underlying molecular mechanisms remains a topic of scientific debate. In this astrochemical context, we present a combined theoretical study using ab initio molecular dynamics (AIMD) based on density functional theory (DFT) and machine learning potentials (PML) constructed with artificial neural networks (ANN) to study the final part of the DIET mechanism in amorphous CO ices. Here, a highly vibrationally excited CO molecule (v = 40) at the center of an aggregate initially composed of 50 CO molecules, optimized and then thermalized at 15 K, triggers the indirect desorption of surface molecules. Our theoretical results reveal that the desorption process consists of three fundamental steps, beginning with a mutual attraction between the vibrationally excited molecule and one or two neighboring molecules, activated by CO bond stretching and facilitated by the steric effect of surrounding molecules. This is followed by a sequence of energy transfers initiated by a collision, resulting in the desorption of vibrationally cold CO molecules in 88% of the AIMD trajectories. Additionally, the theoretical distributions of the internal and translational energy of desorbed molecules remarkably match experimental results, supporting the crucial role of vibrational relaxation in the desorption process. Finally, the first PML constructed from AIMD simulations accurately fit the multidimensional potential energy surface of the system, allowing efficient prediction of aggregate energies and atomic forces. Classical molecular dynamics simulations using these potentials are over 1800 times faster than those based on AIMD while offering precision comparable to DFT
7
Veit, Max David. "Designing a machine learning potential for molecular simulation of liquid alkanes". Thesis, University of Cambridge, 2019. https://www.repository.cam.ac.uk/handle/1810/290295.
Texto completo da fonteResumo:
Molecular simulation is applied to understanding the behaviour of alkane liquids with the eventual goal of being able to predict the viscosity of an arbitrary alkane mixture from first principles. Such prediction would have numerous scientific and industrial applications, as alkanes are the largest component of fuels, lubricants, and waxes; furthermore, they form the backbones of a myriad of organic compounds. This dissertation details the creation of a potential, a model for how the atoms and molecules in the simulation interact, based on a systematic approximation of the quantum mechanical potential energy surface using machine learning. This approximation has the advantage of producing forces and energies of nearly quantum mechanical accuracy at a tiny fraction of the usual cost. It enables accurate simulation of the large systems and long timescales required for accurate prediction of properties such as the density and viscosity. The approach is developed and tested on methane, the simplest alkane, and investigations are made into potentials for longer, more complex alkanes. The results show that the approach is promising and should be pursued further to create an accurate machine learning potential for the alkanes. It could even be extended to more complex molecular liquids in the future.
8
Lundberg, Oscar, Oskar Bjersing e Martin Eriksson. "Approximation of ab initio potentials of carbon nanomaterials with machine learning". Thesis, Luleå tekniska universitet, Institutionen för teknikvetenskap och matematik, 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:ltu:diva-62568.
Texto completo da fonteResumo:
In this work potentials of carbon nanomaterials calculated with Density Functional Theory (DFT) are approximated using an Artificial Neural Network (ANN). Previous work in this field has focused on estimating potential energies of bulk structures. We investigate the possibility to approximate both the potential energies and the forces of periodic carbon nanotubes (CNTs) and fullerenes. The results indicate that for test structures similar to those in the training set the ANN approximates the energies to within 270 meV/atom (< 3.7% error, RMSE 40 meV/atom) and the forces to within 7.5 eV/Å (< 73% error, RMSE 1.34 eV/Å) per atom compared with DFT calculations. Furthermore, we investigate how well the ANN approximates the potentials and forces in structures that are combinations of CNTs and fullerenes (capped CNTs) and find that the ANN generalizes the potential energies to within 100 meV/atom (< 1.1% error, RMSE 78 meV/atom) and the forces to within 6 eV/Å (< 60% error, RMSE 0.55 eV/Å) per atom. The ANN approximated potentials and forces are used to geometry optimize CNTs and we observe that the optimized periodic CNTs match DFT calculated structures and energies while the capped CNTs result in comparable energies but incorrect structures compared to DFT calculations. Considering geometry optimization performed with ANN on CNTs the errors lie within 170 meV/atom (< 1.8% error) with an RMSE of 20 meV/atom. For the geometry optimizations of the capped CNTs the errors are within 430 meV/atom (< 5.5% error) with an RMSE of 14 meV/atom. All results are compared with empirical potentials (ReaxFF) and we find that the ANN approximated potentials are more accurate than the best tested empirical potential. This work shows that machine learning may be used to approximate DFT calculations. However, for further applications our conclusion is that the error of the estimated forces must be reduced further. Finally, we investigate the computing time (number of core hours) required and find that the ANN is about two orders of magnitude faster than DFT and three to four orders of magnitude slower than ReaxFF. For the unseen data the ANN is still around 2 orders of magnitude quicker than the DFT but here it is around 4 order of magnitude slower than ReaxFF.Supervisors: Daniel Hedman and Fredrik Sandin
F7042T - Project in Engineering Physics
9
DRAGONI, DANIELE. "Energetics and thermodynamics of α-iron from first-principles and machine-learning potentials". Doctoral thesis, École Polytechnique Fédérale de Lausanne, 2016. http://hdl.handle.net/10281/231122.
Texto completo da fonteResumo:
Iron is a material of fundamental importance in the industrial and economic processes of our society as it is the major constituent of steels. With advances in computational science, much progress has been made in the understanding of the microscopic mechanisms that determine the macroscopic properties of such material at ordinary or extreme conditions. Ab initio quantum mechanical calculations based on density-functional theory (DFT), in particular, proved to be a unique tool for this purpose. Nevertheless, in order to study large enough systems up to length- and time-scales comparable with those accessible in experiments, interatomic potentials are needed. These are typically based on functional forms driven by physical intuition and fitted on experimental data at zero/low temperature and/or on available first-principles data. Despite their vast success, however, their low flexibility limits their systematic improvement upon database extension. Moreover, their accuracy at intermediate and high temperature remains questionable. In this thesis, we first survey a selection of embedded atom method (EAM) potentials to understand their strengths and limitations in reproducing experimental thermodynamic, vibrational and elastic properties of bcc iron at finite temperature. Our calculations show that, on average, all the potentials rapidly deviate from experiments as temperature is increased. At the same time, they suggest that, despite an anomalous rapid softening of its C44 shear constant, the Mendelev03 parameterization is the most accurate among those considered in this work. As a second step, we compute the same finite-temperature properties from DFT. We verify our plane-wave spin-polarized pseudopotential implementation against selected zero temperature all-electron calculations, thus highlighting the difficulties of the semi-local generalized gradient approximation exchange and correlation functional in describing the electronic properties of iron. On the other hand, we demonstrate that after accounting for the vibrational degrees of freedom, DFT provides a good description of the thermal behavior of thermodynamic and elastic properties of α-iron up to a good fraction of the Curie temperature without the explicit inclusion of magnetic transverse degrees of freedom. Electronic entropy effects are also analyzed and shown to be of secondary importance. Finally, we attempt at generating a set of highly flexible Gaussian approximation potentials (GAP) for bcc iron that retain ab initio accuracy both at zero and finite temperature. To this end, we use a non-linear, non-parametric Gaussian-process regression, and construct a training database of total energies, stresses and forces taken from first-principles molecular dynamics simulations. We cover approximately 105 local atomic environments including pristine and defected bulk systems, and surfaces with different crystallographic orientations. We then validate the different GAP models against DFT data not directly included in the dataset, focusing on the prediction of thermodynamic, vibrational, and elastic properties and of the energetics of bulk defects.
10
Hellsing, Edvin, e Joel Klingberg. "It’s a Match: Predicting Potential Buyers of Commercial Real Estate Using Machine Learning". Thesis, Uppsala universitet, Institutionen för informatik och media, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-445229.
Texto completo da fonteResumo:
This thesis has explored the development and potential effects of an intelligent decision support system (IDSS) to predict potential buyers for commercial real estate property. The overarching need for an IDSS of this type has been identified exists due to information overload, which the IDSS aims to reduce. By shortening the time needed to process data, time can be allocated to make sense of the environment with colleagues. The system architecture explored consisted of clustering commercial real estate buyers into groups based on their characteristics, and training a prediction model on historical transaction data from the Swedish market from the cadastral and land registration authority. The prediction model was trained to predict which out of the cluster groups most likely will buy a given property. For the clustering, three different clustering algorithms were used and evaluated, one density based, one centroid based and one hierarchical based. The best performing clustering model was the centroid based (K-means). For the predictions, three supervised Machine learning algorithms were used and evaluated. The different algorithms used were Naive Bayes, Random Forests and Support Vector Machines. The model based on Random Forests performed the best, with an accuracy of 99.9%.Denna uppsats har undersökt utvecklingen av och potentiella effekter med ett intelligent beslutsstödssystem (IDSS) för att prediktera potentiella köpare av kommersiella fastigheter. Det övergripande behovet av ett sådant system har identifierats existerar på grund av informtaionsöverflöd, vilket systemet avser att reducera. Genom att förkorta bearbetningstiden av data kan tid allokeras till att skapa förståelse av omvärlden med kollegor. Systemarkitekturen som undersöktes bestod av att gruppera köpare av kommersiella fastigheter i kluster baserat på deras köparegenskaper, och sedan träna en prediktionsmodell på historiska transkationsdata från den svenska fastighetsmarknaden från Lantmäteriet. Prediktionsmodellen tränades på att prediktera vilken av grupperna som mest sannolikt kommer köpa en given fastighet. Tre olika klusteralgoritmer användes och utvärderades för grupperingen, en densitetsbaserad, en centroidbaserad och en hierarkiskt baserad. Den som presterade bäst var var den centroidbaserade (K-means). Tre övervakade maskininlärningsalgoritmer användes och utvärderades för prediktionerna. Dessa var Naive Bayes, Random Forests och Support Vector Machines. Modellen baserad p ̊a Random Forests presterade bäst, med en noggrannhet om 99,9%.
11
Ntsaluba, Kuselo Ntsika. "AI/Machine learning approach to identifying potential statistical arbitrage opportunities with FX and Bitcoin Markets". Master's thesis, Faculty of Commerce, 2019. http://hdl.handle.net/11427/31185.
Texto completo da fonteResumo:
In this study, a methodology is presented where a hybrid system combining an evolutionary algorithm with artificial neural networks (ANNs) is designed to make weekly directional change forecasts on the USD by inferring a prediction using closing spot rates of three currency pairs: EUR/USD, GBP/USD and CHF/USD. The forecasts made by the genetically trained ANN are compared to those made by a new variation of the simple moving average (MA) trading strategy, tailored to the methodology, as well as a random model. The same process is then repeated for the three major cryptocurrencies namely: BTC/USD, ETH/USD and XRP/USD. The overall prediction accuracy, uptrend and downtrend prediction accuracy is analyzed for all three methods within the fiat currency as well as the cryptocurrency contexts. The best models are then evaluated in terms of their ability to convert predictive accuracy to a profitable investment given an initial investment. The best model was found to be the hybrid model on the basis of overall prediction accuracy and accrued returns.
12
Sun, Roger. "Utilisation de méthodes radiomiques pour la prédiction des réponses à l’immunothérapie et combinaisons de radioimmunothérapie chez des patients atteints de cancers Radiomics to Assess Tumor Infiltrating CD8 T-Cells and Response to Anti-PD-1/PD-L1 Immunotherapy in Cancer Patients: An Imaging Biomarker Multi-Cohort Study Imagerie médicale computationnelle (radiomique) et potentiel en immuno-oncologie Radiomics to Predict Outcomes and Abscopal Response of Cancer Patients Treated with Immunotherapy Combined with Radiotherapy Using a Validated Signature of CD8 Cells". Thesis, université Paris-Saclay, 2020. http://www.theses.fr/2020UPASL023.
Texto completo da fonteResumo:
Depuis l’arrivée des inhibiteurs de points de contrôle immunitaire, l’immunothérapie a profondément modifié la prise en charge de nombreux cancers, permettant parfois des réponses tumorales prolongées chez des patients atteints de cancers aux stades très avancés. Cependant, malgré des progrès thérapeutiques constants et des associations de traitements combinant par exemple radiothérapie et immunothérapie, la majorité des patients traités ne présentent pas de bénéfices à ces traitements. Ceci explique l’importance de la recherche de biomarqueurs innovants de réponse à l’immunothérapie.L’application de l’intelligence artificielle en imagerie est une discipline récente et en pleine expansion. L’analyse informatique de l’image, appelée également radiomique, permet d’extraire des images médicales de l’information non exploitable à l’œil nu, potentiellement représentative de l’architecture des tissus sous-jacents et de leur composition biologique et cellulaire, et ainsi de développer des biomarqueurs grâce à l’apprentissage automatique (« machine learning »). Cette approche permettrait d’évaluer de façon non invasive la maladie tumorale dans sa globalité, avec la possibilité d’être répétée facilement dans le temps pour appréhender les modifications tumorales survenant au cours de l’histoire de la maladie et de la séquence thérapeutique.Dans le cadre de cette thèse, nous avons évalué si une approche radiomique permettait d’évaluer l’infiltration tumorale lymphocytaire, et pouvait être associée à la réponse de patients traités par immunothérapie. Dans un deuxième temps, nous avons évalué si cette signature permettait d’évaluer la réponse clinique de patients traités par radiothérapie et immunothérapie, et dans quelle mesure elle pouvait être utilisée pour évaluer l’hétérogénéité spatiale tumorale. Les défis spécifiques posés par les données d’imagerie de haute dimension dans le développement d’outils prédictifs applicables en clinique sont discutés dans cette thèseWith the advent of immune checkpoint inhibitors, immunotherapy has profoundly changed the therapeutic strategy of many cancers. However, despite constant therapeutic progress and combinations of treatments such as radiotherapy and immunotherapy, the majority of patients treated do not benefit from these treatments. This explains the importance of research into innovative biomarkers of response to immunotherapyComputational medical imaging, known as radiomics, analyzes and translates medical images into quantitative data with the assumption that imaging reflects not only tissue architecture, but also cellular and molecular composition. This allows an in-depth characterization of tumors, with the advantage of being non-invasive allowing evaluation of tumor and its microenvironment, spatial heterogeneity characterization and longitudinal assessment of disease evolution.Here, we evaluated whether a radiomic approach could be used to assess tumor infiltrating lymphocytes and whether it could be associated with the response of patients treated with immunotherapy. In a second step, we evaluated the association of this radiomic signature with clinical response of patients treated with radiotherapy and immunotherapy, and we assessed whether it could be used to assess tumor spatial heterogeneity.The specific challenges raised by high-dimensional imaging data in the development of clinically applicable predictive tools are discussed in this thesis
13
Syed, Tahir Qasim. "Analysis of the migratory potential of cancerous cells by image preprocessing, segmentation and classification". Thesis, Evry-Val d'Essonne, 2011. http://www.theses.fr/2011EVRY0041/document.
Texto completo da fonteResumo:
Ce travail de thèse s’insère dans un projet de recherche plus global dont l’objectif est d’analyser le potentiel migratoire de cellules cancéreuses. Dans le cadre de ce doctorat, on s’intéresse à l’utilisation du traitement des images pour dénombrer et classifier les cellules présentes dans une image acquise via un microscope. Les partenaires biologistes de ce projet étudient l’influence de l’environnement sur le comportement migratoire de cellules cancéreuses à partir de cultures cellulaires pratiquées sur différentes lignées de cellules cancéreuses. Le traitement d’images biologiques a déjà donné lieu `a un nombre important de publications mais, dans le cas abordé ici et dans la mesure où le protocole d’acquisition des images acquises n'était pas figé, le défi a été de proposer une chaîne de traitements adaptatifs ne contraignant pas les biologistes dans leurs travaux de recherche. Quatre étapes sont détaillées dans ce mémoire. La première porte sur la définition des prétraitements permettant d’homogénéiser les conditions d’acquisition. Le choix d’exploiter l’image des écarts-type plutôt que la luminosité est un des résultats issus de cette première partie. La deuxième étape consiste à compter le nombre de cellules présentent dans l’image. Un filtre original, nommé filtre «halo», permettant de renforcer le centre des cellules afin d’en faciliter leur comptage, a été proposé. Une étape de validation statistique de ces centres permet de fiabiliser le résultat obtenu. L’étape de segmentation des images, sans conteste la plus difficile, constitue la troisième partie de ce travail. Il s’agit ici d’extraire des «vignettes», contenant une seule cellule. Le choix de l’algorithme de segmentation a été celui de la «Ligne de Partage des Eaux», mais il a fallu adapter cet algorithme au contexte des images faisant l’objet de cette étude. La proposition d’utiliser une carte de probabilités comme données d’entrée a permis d’obtenir une segmentation au plus près des bords des cellules. Par contre cette méthode entraine une sur-segmentation qu’il faut réduire afin de tendre vers l’objectif : «une région = une cellule». Pour cela un algorithme utilisant un concept de hiérarchie cumulative basée morphologie mathématique a été développée. Il permet d’agréger des régions voisines en travaillant sur une représentation arborescente de ces régions et de leur niveau associé. La comparaison des résultats obtenus par cette méthode à ceux proposés par d’autres approches permettant de limiter la sur-segmentation a permis de prouver l’efficacité de l’approche proposée. L’étape ultime de ce travail consiste dans la classification des cellules. Trois classes ont été définies : cellules allongées (migration mésenchymateuse), cellules rondes «blebbantes» (migration amiboïde) et cellules rondes «lisses» (stade intermédiaire du mode de migration). Sur chaque vignette obtenue à la fin de l’étape de segmentation, des caractéristiques de luminosité, morphologiques et texturales ont été calculées. Une première analyse de ces caractéristiques a permis d’élaborer une stratégie de classification, à savoir séparer dans un premier temps les cellules rondes des cellules allongées, puis séparer les cellules rondes «lisses» des «blebbantes». Pour cela on divise les paramètres en deux jeux qui vont être utilisés successivement dans ces deux étapes de classification. Plusieurs algorithmes de classification ont été testés pour retenir, au final, l’utilisation de deux réseaux de neurones permettant d’obtenir plus de 80% de bonne classification entre cellules longues et cellules rondes, et près de 90% de bonne classification entre cellules rondes «lisses» et «blebbantes»This thesis is part of a broader research project which aims to analyze the potential migration of cancer cells. As part of this doctorate, we are interested in the use of image processing to count and classify cells present in an image acquired usinga microscope. The partner biologists of this project study the influence of the environment on the migratory behavior of cancer cells from cell cultures grown on different cancer cell lines. The processing of biological images has so far resulted in a significant number of publications, but in the case discussed here, since the protocol for the acquisition of images acquired was not fixed, the challenge was to propose a chain of adaptive processing that does not constrain the biologists in their research. Four steps are detailed in this paper. The first concerns the definition of pre-processing steps to homogenize the conditions of acquisition. The choice to use the image of standard deviations rather than the brightness is one of the results of this first part. The second step is to count the number of cells present in the image. An original filter, the so-called “halo” filter, that reinforces the centre of the cells in order to facilitate counting, has been proposed. A statistical validation step of the centres affords more reliability to the result. The stage of image segmentation, undoubtedly the most difficult, constitutes the third part of this work. This is a matter of extracting images each containing a single cell. The choice of segmentation algorithm was that of the “watershed”, but it was necessary to adapt this algorithm to the context of images included in this study. The proposal to use a map of probabilities as input yielded a segmentation closer to the edges of cells. As against this method leads to an over-segmentation must be reduced in order to move towards the goal: “one region = one cell”. For this algorithm the concept of using a cumulative hierarchy based on mathematical morphology has been developed. It allows the aggregation of adjacent regions by working on a tree representation ofthese regions and their associated level. A comparison of the results obtained by this method with those proposed by other approaches to limit over-segmentation has allowed us to prove the effectiveness of the proposed approach. The final step of this work consists in the classification of cells. Three classes were identified: spread cells (mesenchymal migration), “blebbing” round cells (amoeboid migration) and “smooth” round cells (intermediate stage of the migration modes). On each imagette obtained at the end of the segmentation step, intensity, morphological and textural features were calculated. An initial analysis of these features has allowed us to develop a classification strategy, namely to first separate the round cells from spread cells, and then separate the “smooth” and “blebbing” round cells. For this we divide the parameters into two sets that will be used successively in Two the stages of classification. Several classification algorithms were tested, to retain in the end, the use of two neural networks to obtain over 80% of good classification between long cells and round cells, and nearly 90% of good Classification between “smooth” and “blebbing” round cells
14
Egieyeh, Samuel Ayodele. "Computational strategies to identify, prioritize and design potential antimalarial agents from natural products". University of the Western Cape, 2015. http://hdl.handle.net/11394/5058.
Texto completo da fonteResumo:
Philosophiae Doctor - PhDIntroduction: There is an exigent need to develop novel antimalarial drugs in view of the mounting disease burden and emergent resistance to the presently used drugs against the malarial parasites. A large amount of natural products, especially those used in ethnomedicine for malaria, have shown varying in-vitro antiplasmodial activities. Facilitating antimalarial drug development from this wealth of natural products is an imperative and laudable mission to pursue. However, the limited resources, high cost, low prospect and the high cost of failure during preclinical and clinical studies might militate against pursue of this mission. Chemoinformatics techniques can simulate and predict essential molecular properties required to characterize compounds thus eliminating the cost of equipment and reagents to conduct essential preclinical studies, especially on compounds that may fail during drug development. Therefore, applying chemoinformatics techniques on natural products with in-vitro antiplasmodial activities may facilitate identification and prioritization of these natural products with potential for novel mechanism of action, desirable pharmacokinetics and high likelihood for development into antimalarial drugs. In addition, unique structural features mined from these natural products may be templates to design new potential antimalarial compounds. Method: Four chemoinformatics techniques were applied on a collection of selected natural products with in-vitro antiplasmodial activity (NAA) and currently registered antimalarial drugs (CRAD): molecular property profiling, molecular scaffold analysis, machine learning and design of a virtual compound library. Molecular property profiling included computation of key molecular descriptors, physicochemical properties, molecular similarity analysis, estimation of drug-likeness, in-silico pharmacokinetic profiling and exploration of structure-activity landscape. Analysis of variance was used to assess statistical significant differences in these parameters between NAA and CRAD. Next, molecular scaffold exploration and diversity analyses were performed on three datasets (NAA, CRAD and malarial data from Medicines for Malarial Ventures (MMV)) using scaffold counts and cumulative scaffold frequency plots. Scaffolds from the NAA were compared to those from CRAD and MMV. A Scaffold Tree was also generated for all the datasets. Thirdly, machine learning approaches were used to build four regression and four classifier models from bioactivity data of NAA using molecular descriptors and molecular fingerprints. Models were built and refined by leave-one-out cross-validation and evaluated with an independent test dataset. Applicability domain (AD), which defines the limit of reliable predictability by the models, was estimated from the training dataset and validated with the test dataset. Possible chemical features associated with reported antimalarial activities of the compounds were also extracted. Lastly, virtual compound libraries were generated with the unique molecular scaffolds identified from the NAA. The virtual compounds generated were characterized by evaluating selected molecular descriptors, toxicity profile, structural diversity from CRAD and prediction of antiplasmodial activity. Results: From the molecular property profiling, a total of 1040 natural products were selected and a total of 13 molecular descriptors were analyzed. Significant differences were observed between the natural products with in-vitro antiplasmodial activities (NAA) and currently registered antimalarial drugs (CRAD) for at least 11 of the molecular descriptors. Molecular similarity and chemical space analysis identified NAA that were structurally diverse from CRAD. Over 50% of NAA with desirable drug-like properties were identified. However, nearly 70% of NAA were identified as potentially "promiscuous" compounds. Structure-activity landscape analysis highlighted compound pairs that formed "activity cliffs". In all, prioritization strategies for the natural products with in-vitro antiplasmodial activities were proposed. The scaffold exploration and analysis results revealed that CRAD exhibited greater scaffold diversity, followed by NAA and MMV respectively. Unique scaffolds that were not contained in any other compounds in the CRAD datasets were identified in NAA. The Scaffold Tree showed the preponderance of ring systems in NAA and identified virtual scaffolds, which maybe potential bioactive compounds or elucidate the NAA possible synthetic routes. From the machine learning study, the regression and classifier models that were most suitable for NAA were identified as model tree M5P (correlation coefficient = 0.84) and Sequential Minimization Optimization (accuracy = 73.46%) respectively. The test dataset fitted into the applicability domain (AD) defined by the training dataset. The “amine” group was observed to be essential for antimalarial activity in both NAA and MMV dataset but hydroxyl and carbonyl groups may also be relevant in the NAA dataset. The results of the characterization of the virtual compound library showed significant difference (p value < 0.05) between the virtual compound library and currently registered antimalarial drugs in some molecular descriptors (molecular weight, log partition coefficient, hydrogen bond donors and acceptors, polar surface area, shape index, chiral centres, and synthetic feasibility). Tumorigenic and mutagenic substructures were not observed in a large proportion (> 90%) of the virtual compound library. The virtual compound libraries showed sufficient diversity in structures and majority were structurally diverse from currently registered antimalarial drugs. Finally, up to 70% of the virtual compounds were predicted as active antiplasmodial agents. Conclusions:Molecular property profiling of natural products with in-vitro antiplasmodial activities (NAA) and currently registered antimalarial drugs (CRAD) produced a wealth of information that may guide decisions and facilitate antimalarial drug development from natural products and led to a prioritized list of natural products with in-vitro antiplasmodial activities. Molecular scaffold analysis identified unique scaffolds and virtual scaffolds from NAA that possess desirable drug-like properties, which make them ideal starting points for molecular antimalarial drug design. The machine learning study built, evaluated and identified amply accurate regression and classifier accurate models that were used for virtual screening of natural compound libraries to mine possible antimalarial compounds without the expense of bioactivity assays. Finally, a good amount of the virtual compounds generated were structurally diverse from currently registered antimalarial drugs and potentially active antiplasmodial agents. Filtering and optimization may lead to a collection of virtual compounds with unique chemotypes that may be synthesized and added to screening deck against Plasmodium.
15
Skabar, Andrew Alojz. "Inductive learning techniques for mineral potential mapping". Thesis, Queensland University of Technology, 2001.
Encontre o texto completo da fonte
16
Gayraud, Nathalie. "Méthodes adaptatives d'apprentissage pour des interfaces cerveau-ordinateur basées sur les potentiels évoqués". Thesis, Université Côte d'Azur (ComUE), 2018. http://www.theses.fr/2018AZUR4231/document.
Texto completo da fonteResumo:
Les interfaces cerveau machine (BCI pour Brain Computer Interfaces) non invasives permettent à leur utilisateur de contrôler une machine par la pensée. Ce dernier doit porter un dispositif d'acquisition de signaux électroencéphalographiques (EEG), qui sont dotés d'un rapport signal sur bruit assez faible ; à ceci s'ajoute l’importante variabilité tant à travers les sessions d'utilisation qu’à travers les utilisateurs. Par conséquent, la calibration du BCI est souvent nécessaire avant son utilisation. Cette thèse étudie les sources de cette variabilité, dans le but d'explorer, concevoir, et implémenter des méthodes d'autocalibration. Nous étudions la variabilité des potentiels évoqués, particulièrement une composante tardive appelée P300. Nous nous penchons sur trois méthodes d’apprentissage par transfert : la Géométrie Riemannienne, le Transport Optimal, et l’apprentissage ensembliste. Nous proposons un modèle de l'EEG qui tient compte de la variabilité. Les paramètres résultants de nos analyses nous servent à calibrer ce modèle et à simuler une base de données, qui nous sert à évaluer la performance des méthodes d’apprentissage par transfert. Puis ces méthodes sont combinées et appliquées à des données expérimentales. Nous proposons une méthode de classification basée sur le Transport Optimal dont nous évaluons la performance. Ensuite, nous introduisons un marqueur de séparabilité qui nous permet de combiner Géométrie Riemannienne, Transport Optimal et apprentissage ensembliste. La combinaison de plusieurs méthodes d’apprentissage par transfert nous permet d’obtenir un classifieur qui s’affranchit des différentes sources de variabilité des signaux EEGNon-invasive Brain Computer Interfaces (BCIs) allow a user to control a machine using only their brain activity. The BCI system acquires electroencephalographic (EEG) signals, characterized by a low signal-to-noise ratio and an important variability both across sessions and across users. Typically, the BCI system is calibrated before each use, in a process during which the user has to perform a predefined task. This thesis studies of the sources of this variability, with the aim of exploring, designing, and implementing zero-calibration methods. We review the variability of the event related potentials (ERP), focusing mostly on a late component known as the P300. This allows us to quantify the sources of EEG signal variability. Our solution to tackle this variability is to focus on adaptive machine learning methods. We focus on three transfer learning methods: Riemannian Geometry, Optimal Transport, and Ensemble Learning. We propose a model of the EEG takes variability into account. The parameters resulting from our analyses allow us to calibrate this model in a set of simulations, which we use to evaluate the performance of the aforementioned transfer learning methods. These methods are combined and applied to experimental data. We first propose a classification method based on Optimal Transport. Then, we introduce a separability marker which we use to combine Riemannian Geometry, Optimal Transport and Ensemble Learning. Our results demonstrate that the combination of several transfer learning methods produces a classifier that efficiently handles multiple sources of EEG signal variability
17
Nyman, Måns, e Caner Naim Ulug. "Exploring the Potential for Machine Learning Techniques to Aid in Categorizing Electron Trajectories during Magnetic Reconnection". Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-279982.
Texto completo da fonteResumo:
Magnetic reconnection determines the space weather which has a direct impact on our contemporary technological systems. As such, the phenomenon has serious ramifications on humans. Magnetic reconnection is a topic which has been studied for a long time, yet still many aspects surrounding the phenomenon remain unexplored. Scientists within the field believe that the electron dynamics play an important role in magnetic reconnection. During magnetic reconnection, electrons can be accelerated to high velocities. A large number of studies have been made regarding the trajectories that these electrons exhibit and researchers in this field could easily point out what type of trajectory a specific electron exhibits given a plot of said trajectory. Attempting to do this for a more realistic number of electrons manually is however not an easy or efficient task to take on. By using Machine Learning techniques to attempt to categorize these trajectories, this process could be sped up immensely. Yet to date there has been no attempt at this. In this thesis, an attempt to answer how certain Machine Learning techniques perform in this matter was made. Principal component analysis and K-means clustering were the main methods applied after using different preprocessing methods on the given data set. The Elbow method was employed to find the optimal K-value and was complemented by Self-Organizing Maps. Silhouette coefficient was used to measure the performance of the methods. The First-centering and Mean-centering preprocessing methods yielded the two highest silhouette coefficients, thus displaying the best quantitative performances. However, inspection of the clusters pointed to a lack of perfect overlap between the classes detected by employed techniques and the classes identified in previous physics articles. Nevertheless, Machine Learning methods proved to possess certain potential that is worth exploring in greater detail in future studies in the field of magnetic reconnection.Magnetisk rekonnektion påverkar rymdvädret som har en direkt påverkan på våra nutida teknologiska system. Således kan fenomenet ge allvarliga konsekvenser för människor. Forskare inom detta fält tror att elektrondynamiken spelar en viktig roll i magnetisk rekonnektion. Magnetisk rekonnektion är ett ämne som har studerats under lång tid men ännu förblir många aspekter av fenomenet outforskade. Under magnetisk rekonnektion kan elektroner accelereras till höga hastigheter. En stor mängd studier har gjorts angående trajektorierna som dessa elektroner uppvisar och forskare som är aktiva inom detta forskningsområde skulle enkelt kunna bestämma vilken sorts trajektoria en specifik elektron uppvisar givet en grafisk illustration av sagda trajektoria. Att försöka göra detta för ett mer realistiskt antal elektroner manuellt är dock ingen enkel eller effektiv uppgift att ta sig an. Genom användning av Maskininlärningstekniker för att försöka kategorisera dessa trajektorier skulle denna process kunna göras mycket mer effektiv. Ännu har dock inga försök att göra detta gjorts. I denna uppsats gjordes ett försök att besvara hur väl vissa Maskinlärningstekniker presterar i detta avseende. Principal component analysis och K-means clustering var huvudmetoderna som användes, applicerade med olika sorters förbehandling av den givna datan. Elbow-metoden användes för att hitta det optimala K-värdet och kompletterades av Self-Organizing Maps. Silhouette coefficient användes för att mäta resultaten av dessa metoder. Förbehandlingsmetoderna First-centering och Mean-centering gav de två högsta siluett-koefficienterna och uppvisade således de bästa kvantitativa resultaten. Inspektion av klustrarna pekade dock på avsaknad av perfekt överlappning, både mellan klasserna som upptäcktes av de tillämpade metoderna samt klasserna som har identifierats i tidigare artiklar inom fysik. Trots detta visade sig Maskininlärningsmetoder besitta viss potential som är värt att utforska i större detalj i framtida studier inom fältet magnetisk rekonnektion.
18
Zaremba, Wojciech. "Modeling the variability of EEG/MEG data through statistical machine learning". Habilitation à diriger des recherches, Ecole Polytechnique X, 2012. http://tel.archives-ouvertes.fr/tel-00803958.
Texto completo da fonteResumo:
Brain neural activity generates electrical discharges, which manifest as electrical and magnetic potentials around the scalp. Those potentials can be registered with magnetoencephalography (MEG) and electroencephalography (EEG) devices. Data acquired by M/EEG is extremely difficult to work with due to the inherent complexity of underlying brain processes and low signal-to-noise ratio (SNR). Machine learning techniques have to be employed in order to reveal the underlying structure of the signal and to understand the brain state. This thesis explores a diverse range of machine learning techniques which model the structure of M/EEG data in order to decode the mental state. It focuses on measuring a subject's variability and on modeling intrasubject variability. We propose to measure subject variability with a spectral clustering setup. Further, we extend this approach to a unified classification framework based on Laplacian regularized support vector machine (SVM). We solve the issue of intrasubject variability by employing a model with latent variables (based on a latent SVM). Latent variables describe transformations that map samples into a comparable state. We focus mainly on intrasubject experiments to model temporal misalignment.
19
Evett, Chantal. "What are the Potential Impacts of Big Data, Artificial Intelligence and Machine Learning on the Auditing Profession?" Scholarship @ Claremont, 2017. http://scholarship.claremont.edu/cmc_theses/1642.
Texto completo da fonteResumo:
To maintain public confidence in the financial system, it is essential that most financial fraud is prevented and that incidents of fraud are detected and punished. The responsibility of uncovering creatively implemented fraud is placed, in a large part, on auditors. Recent advancements in technology are helping auditors turn the tide against fraudsters. Big Data, made possible by the proliferation, widespread availability and amalgamation of diverse digital data sets, has become an important driver of technological change. Big Data analytics are already transforming the traditional audit. Sampling and testing a limited number of random samples has turned into a much more comprehensive audit that analyzes the entire population of transactions within an account, allowing auditors to flag and investigate all sorts of potentially fraudulent anomalies that were previously invisible. Artificial intelligence (AI) programs, typified by IBM’s Watson, can mimic the thought processes of the human mind and will soon be adopted by the auditing profession. Machine learning (ML) programs, with the ability to change when exposed to new data, are developing rapidly and may take over many of the decision-making functions currently performed by auditors. The SEC has already implemented pioneering fraud-detection software based on AI and ML programs. The evolution of the auditor’s role has already begun. Current accounting students must understand the traditional auditing skillset will not longer be sufficient. While facing a future with fewer auditing positions available due to increased automation, auditors will need training for roles that will be more data analytical and computer-science based.
20
Kottorp, Max, e Filip Jäderberg. "Chatbot As a Potential Tool for Businesses : A study on chatbots made in collaboration with Bisnode". Thesis, KTH, Industriell ekonomi och organisation (Inst.), 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-210768.
Texto completo da fonteResumo:
The investigation aims to provide an answer to if a chatbot is a potential complement to an internal service desk of a company. The work has centered around developing a chatbot able to handle simple Q&A-interaction of the internal service desk of Bisnode, the company in question. The chatbot acted as an proof of concept, which then was tested by 15 individuals. The testing was done with pre- defined user scenarios, where the test person ultimately had to fill in a questionnaire with statements related to the overall experience. By summarizing the user evaluations from the questionnaires, combined with an SWOT analysis, the work concluded that a chatbot is indeed a potential complement to an internal service desk of a company, if it handles Q&A-interaction.
21
Li, Zhenwei. "On-the-fly machine learning of quantum mechanical forces and its potential applications for large scale molecular dynamics". Thesis, King's College London (University of London), 2014. http://kclpure.kcl.ac.uk/portal/en/theses/onthefly-machine-learning-of-quantum-mechanical-forces-and-its-potential-applications-for-large-scale-molecular-dynamics(2a2f33a6-fa0c-44e3-8689-f4cf3f1c9198).html.
Texto completo da fonteResumo:
Material simulation using molecular dynamics (MD) at the quantum mechanical (QM) accuracy level has gained great interest in the community. However, the bottleneck arising from the O(N3) scaling of QM calculation has enormously limited its investigation scope. As an approach to address this issue, in this thesis, I proposed a machine-learning (ML) MD scheme based on Bayesian inference from CPU-intensive QM force database. In this scheme, QM calculations are only performed when necessary and used to augment the ML database for more challenging prediction case. The scheme is generally transferable to new chemical situations and database completeness is never required. To achieve the maximal ML eciency, I use a symmetrically reduced internal-vector representation for the atomic congurations. Signicant speed-up factor is achieved under controllable accuracy tolerance in the MD simulation on test case of Silicon at dierent temperatures. As the database grows in conguration space, the extrapolative capability systematically increases and QM calculations are nally not needed for simple chemical processes. In the on-the-y ML force calculation scheme, sorting/selecting out the closest data congurations is used to enhance the overall eciency to scale as O(N). The potential application of this methodology for large-scale simulation (e.g. fracture, amorphous, defect), where chemical accuracy and computational eciency are required at the same time, can be anticipated. In the context of fracture simulations, a typical multi-scale system, interesting events happen near the crack tips beyond the description of classical potentials. The simulation results by machine-learning potential derived from a xed database with no enforced QM accuracy inspire a theoretical model which is further used to investigate the atomic bond breaking process during fracture propagation as well as its relation with the initialised vibration modes, crack speed, and bonding structure.
22
BALESTRUCCI, ALESSANDRO. "Potential target audience of misinformation on Social Media: Credulous Users". Doctoral thesis, Gran Sasso Science Institute, 2020. http://hdl.handle.net/20.500.12571/14754.
Texto completo da fonteResumo:
The capability to reach a wider audience and the possibility to disseminate news faster are the main reasons for the growing importance of Online Social Media (OSM) whose usage has undoubtedly reshaped the way news are written, published and disseminated.
However, due to the technical limits of online fact-checkers and to an uncontrolled content publishing, there is a high risk of being misinformed through fake news.
Although automated accounts known as bots are considered the main promoters of mis-/dis- information diffusion, those who, with their actions, change the current events (e.g., welfare, economy, politics, etc.) are human users.
Some categories of humans are more vulnerable to fake news than others,
and performing mis-/dis- information activities targeting such categories would increase the efficacy of such activities. Furthermore, recent studies have evidenced users' attitude not to fact-check the news they diffuse on OSM and thus the risk that they became themselves vectors of mis-/dis- information.
In this document, using Twitter as a benchmark, we devote our attention to those human-operated accounts, named ``credulous'' users, which show a relatively high number of bots as followees (called bot-followees).
We believe that these users are more vulnerable to manipulation (w.r.t. other human-operated accounts) and, although unknowingly, they can be involved in malicious activities such as diffusion of fake content. Specifically, we first design some heuristics by focusing on the aspects that best characterise the credulous users w.r.t. not credulous ones. Then, by applying Machine Learning (ML) techniques, we develop an approach based on binary classifiers able to automatically identify this kind of users and then use regression models to predict the percentage of humans' bot-followees (over their respective followees). Subsequently, we describe investigations conducted to ascertain the actual contribution of credulous users in the dissemination of potentially malicious content and then, their involvements in fake news diffusion by analysing and comparing the fake news spread by credulous users w.r.t. not credulous one.
Our investigations and experiments provide evidence of credulous users' involvement in spreading fake news thus supporting bots' actions on OSM.
23
Hagward, Anders. "Using Git Commit History for Change Prediction : An empirical study on the predictive potential of file-level logical coupling". Thesis, KTH, Skolan för datavetenskap och kommunikation (CSC), 2015. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-172998.
Texto completo da fonteResumo:
In recent years, a new generation of distributed version control systems have taken the place of the aging centralized ones, with Git arguably being the most popular distributed system today. We investigate the potential of using Git commit history to predict files that are often changed together. Specifically, we look at the rename tracking heuristic found in Git, and the impact it has on prediction performance. By applying a data mining algorithm to five popular GitHub repositories we extract logical coupling – inter-file dependencies not necessarily detectable by static analysis – on which we base our change prediction. In addition, we examine if certain commits are better suited for change prediction than others; we define a bug fix commit as a commit that resolves one or more issues in the associated issue tracking system and compare their prediction performance. While our findings do not reveal any notable differences in prediction performance when disregarding rename information, they suggest that extracting coupling from, and predicting on, bug fix commits in particular could lead to predictions that are both more accurate and numerous.De senaste åren har en ny generation av distribuerade versionshanteringssystem tagit plats där tidigare centraliserade sådana huserat. I spetsen för dessa nya system går ett system vid namn Git. Vi undersöker potentialen i att nyttja versionshistorik från Git i syftet att förutspå filer som ofta redigeras ihop. I synnerhet synar vi Gits heuristik för att detektera när en fil flyttats eller bytt namn, någonting som torde vara användbart för att bibehålla historiken för en sådan fil, och mäter dess inverkan på prediktionsprestandan. Genom att applicera en datautvinningsalgoritm på fem populära GitHubprojekt extraherar vi logisk koppling – beroenden mellan filer som inte nödvändigtvis är detekterbara medelst statisk analys – på vilken vi baserar vår prediktion. Därtill utreder vi huruvida vissa Gitcommits är bättre lämpade för prediktion än andra; vi definierar en buggfixcommit som en commit som löser en eller flera buggar i den tillhörande buggdatabasen, och jämför deras prediktionsprestanda. Medan våra resultat ej kan påvisa några större prestandamässiga skillnader när flytt- och namnbytesinformationen ignorerades, indikerar de att extrahera koppling från, och prediktera på, enbart bugfixcommits kan leda till förutsägelser som är både mer precisa och mångtaliga.
24
Ioannides, Charalambos. "Investigating the potential of machine learning techniques for feedback-based coverage-directed test genreation in simulation-based digital design verification". Thesis, University of Bristol, 2013. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.618315.
Texto completo da fonteResumo:
A consistent trend in the semiconductor industry has been the increase of embedded functionality in new designs. As a result, the verification process today requires significant resources to cope with these increasingly complex designs. In order to alleviate the problem, industrialists and academics have proposed and improved on many formal, simulation-based and hybrid verification techniques. To dale, none of the approaches proposed have been ab le to present a convincing argument warranting their unconditional adoption by the industry. In an attempt to further automate design verification (DV), especially in simulation-based and hybrid approaches, machine learning (ML) techniques have been exploited to close the loop between coverage feedback and test generation ; a process also known as coverage directed test generation (COG). Although most techniques in the literature are reported to help in constructing minimal tests that exercise most, if not the entire design under verification, a question remains on their practical usefulness when applied in real-world industry-level verification environments. The aim of this work was to answer the following questions: I. What would constitute a good ML-COG solution? What would be its characteristics? 2. 00 existing ML-CDG technique(s) scale to industrial designs and verification environments? 3. Can we develop an ML-based system that can attempt functional coverage balancing? This work answers these questions having gathered requirements and capabilities from earlier academic work and having filtered them through an industrial perspective on usefulness and practicality. The main metrics used to evaluate these were effectiveness in terms of coverage achieved and effort in terms of computation time. A coverage closure effective and easy to use genetic programming technique has been applied on an industrial level verification project and the poor results obtained show that the particular technique does not scale well. Linear regress ion has been attempted for feature extraction as part of a larger and novel stochastic ML-CDG model. The results on the capability of these techniques were again below expectations thus showing the ineffectiveness of these algorithms on larger datasets. Finally, learning classifier systems, specifically XCS, have been used to discover the cause-effect relationships between test generator biases and coverage. The results obtained pointed to a problem with the learning mechanism in XCS, and a misconception held by academics on its capabilities. Though XCS at its current state is not an immediately exploitable ML~CDG technique, it shows the necessary potential for later adoption once the problem discovered here is resolved through further research. The outcome of this research was the realisation that the contemporary ML methodologies that have been experimented with fall short of expectations when dealing with industry-level simulation-based digital design verification. In addition, it was discovered that design verification constitutes a problem area that can stress these techniques to their limits and can therefore indicate areas for further improvement and academic research.
25
Leo, Stephen. "Potential of remote and proximal sensing, publicly available datasets and machine learning for site-specific management in Australian irrigated cotton systems". Thesis, Queensland University of Technology, 2022. https://eprints.qut.edu.au/235383/1/Stephen%2BLeo%2BThesis%281%29.pdf.
Texto completo da fonteResumo:
Agricultural fields are inherently variable across both space and time but are commonly managed uniformly. Uniform management can simultaneously lead to an under and over-application of resources (e.g. fertiliser) within the same field, resulting in poor resource efficiency and reduced profit margins. This research demonstrated the potential of publicly available datasets (i.e. remote sensing, digital soil maps, weather), machine learning techniques and crop models to inform management at a sub-paddock scale. These findings will help provide a cost-effective and efficient approach to improving farm productivity, profitability and sustainability in Australian irrigated cotton systems.
26
Jahangiri, Arash. "Investigating Violation Behavior at Intersections using Intelligent Transportation Systems: A Feasibility Analysis on Vehicle/Bicycle-to-Infrastructure Communications as a Potential Countermeasure". Diss., Virginia Tech, 2015. http://hdl.handle.net/10919/76729.
Texto completo da fonteResumo:
The focus of this dissertation is on safety improvement at intersections and presenting how Vehicle/Bicycle-to-Infrastructure Communications can be a potential countermeasure for crashes resulting from drivers' and cyclists' violations at intersections. The characteristics (e.g., acceleration capabilities, etc.) of transportation modes affect the violation behavior. Therefore, the first building block is to identify the users' transportation mode. Consequently, having the mode information, the second building block is to predict whether or not the user is going to violate. This step focuses on two different modes (i.e., driver violation prediction and cyclist violation prediction). Warnings can then be issued for users in potential danger to react or for the infrastructure and vehicles so they can take appropriate actions to avoid or mitigate crashes.
A smartphone application was developed to collect sensor data used to conduct the transportation mode recognition task. Driver violation prediction task at signalized intersections was conducted using observational and simulator data. Also, a naturalistic cycling experiment was designed for cyclist violation prediction task. Subsequently, cyclist violation behavior was investigated at both signalized and stop-controlled intersections. To build the prediction models in all the aforementioned tasks, various Artificial Intelligence techniques were adopted. K-fold Cross-Validation as well as Out-of-Bag error was used for model selection and validation.
Transportation mode recognition models contributed to high classification accuracies (e.g., up to 98%). Thus, data obtained from the smartphone sensors were found to provide important information to distinguish between transportation modes. Driver violation (i.e., red light running) prediction models were resulted in high accuracies (i.e., up to 99.9%). Time to intersection (TTI), distance to intersection (DTI), the required deceleration parameter (RDP), and velocity at the onset of a yellow light were among the most important factors in violation prediction. Based on logistic regression analysis, movement type and presence of other users were found as significant factors affecting the probability of red light violations by cyclists at signalized intersections. Also, presence of other road users and age were the significant factors affecting violations at stop-controlled intersections. In case of stop-controlled intersections, violation prediction models resulted in error rates of 0 to 10 percent depending on how far from the intersection the prediction task is conducted.Ph. D.
27
Sepp, Löfgren Nicholas. "Accelerating bulk material property prediction using machine learning potentials for molecular dynamics : predicting physical properties of bulk Aluminium and Silicon". Thesis, Linköpings universitet, Teoretisk Fysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-179894.
Texto completo da fonteResumo:
In this project machine learning (ML) interatomic potentials are trained and used in molecular dynamics (MD) simulations to predict the physical properties of total energy, mean squared displacement (MSD) and specific heat capacity for systems of bulk Aluminium and Silicon. The interatomic potentials investigated are potentials trained using the ML models kernel ridge regression (KRR) and moment tensor potentials (MTPs). The simulations using these ML potentials are then compared with results obtained from ab-initio simulations using the gold standard method of density functional theory (DFT), as implemented in the Vienna ab-intio simulation package (VASP). The results show that the MTP simulations reach comparable accuracy compared to the DFT simulations for the properties total energy and MSD for Aluminium, with errors in the orders of magnitudes of meV and 10-5 Å2. Specific heat capacity is not reasonably replicated for Aluminium. The MTP simulations do not reasonably replicate the studied properties for the system of Silicon. The KRR models are implemented in the most direct way, and do not yield reasonably low errors even when trained on all available 10000 time steps of DFT training data. On the other hand, the MTPs require only to be trained on approximately 100 time steps to replicate the physical properties of Aluminium with accuracy comparable to DFT. After being trained on 100 time steps, the trained MTPs achieve mean absolute errors in the orders of magnitudes for the energy per atom and force magnitude predictions of 10-3 and 10-1 respectively for Aluminium, and 10-3 and 10-2 respectively for Silicon. At the same time, the MTP simulations require less core hours to simulate the same amount of time steps as the DFT simulations. In conclusion, MTPs could very likely play a role in accelerating both materials simulations themselves and subsequently the emergence of the data-driven materials design and informatics paradigm.
28
Chowdhury, Ziaul Islam, e Iskanter Bensenousi. "Evaluation of different machine learning models for the prediction of electric or hybrid vehicle buyers and identification of the characteristics of the buyers in the EU". Thesis, Blekinge Tekniska Högskola, Institutionen för industriell ekonomi, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-20712.
Texto completo da fonteResumo:
The main goal of this thesis is to evaluate different machine learning models in order to classify buyers of an electric or a hybrid vehicle and to identify the characteristics of the buyers in the European Union. Machine learning algorithms and techniques were adopted to analyze the dataset and to create models that could predict, with a certain accuracy, the customer’s willingness to buy an EV. Identification of the characteristics of the buyers were based on the identified most important features from the machine learning models and statistical analysis. The research consisted of exploratory and explanatory methods (mixed method) with quantitative and qualitative techniques. Quantitative technique was applied to convert categorical values to ordinal and nominal numeric values, to establish cause and effect relationship between the variables by using statistical analysis and to apply machine learning methods on the dataset. The quantitative results were then analyzed by using quantitative and qualitative techniques in order to identify the characteristics of the buyers. The data analytics part relied on a publicly available large dataset from the EU containing transport and mobility related data. From the experiments with logistics regression, support vector machine, random forest, gradient boosting classifier and the artificial neural network it was found that ANN is the best model to identify who won’t buy an EV and gradient boosting classifier is the best model to identify who would like to buy and EV. ML based feature importance identification methods (MDI, permutation feature importance) were used to analyze the characteristics of the buyers. The major buyer’s characteristics found in this thesis are environmental concern, knowledge on car sharing, country of residence, education, control traffic, gender, incentive, education and location of residence. Authors have recommended green marketing as the potential enablers towards a faster and larger adoption of electrical vehicles in the market as environmental impact was found as the most significant behavior of the buyer. Finally, for the future researchers, the authors have recommended fine-tuning the algorithms extensively in order to achieve better accuracy and to collect primary data based on the most important features identified in this thesis.
29
Zanghieri, Marcello. "sEMG-based hand gesture recognition with deep learning". Master's thesis, Alma Mater Studiorum - Università di Bologna, 2019. http://amslaurea.unibo.it/18112/.
Texto completo da fonteResumo:
Hand gesture recognition based on surface electromyographic (sEMG) signals is a promising approach for the development of Human-Machine Interfaces (HMIs) with a natural control, such as intuitive robot interfaces or poly-articulated prostheses. However, real-world applications are limited by reliability problems due to motion artifacts, postural and temporal variability, and sensor re-positioning.
This master thesis is the first application of deep learning on the Unibo-INAIL dataset, the first public sEMG dataset exploring the variability between subjects, sessions and arm postures, by collecting data over 8 sessions of each of 7 able-bodied subjects executing 6 hand gestures in 4 arm postures. In the most recent studies, the variability is addressed with training strategies based on training set composition, which improve inter-posture and inter-day generalization of classical (i.e. non-deep) machine learning classifiers, among which the RBF-kernel SVM yields the highest accuracy.
The deep architecture realized in this work is a 1d-CNN implemented in Pytorch, inspired by a 2d-CNN reported to perform well on other public benchmark databases. On this 1d-CNN, various training strategies based on training set composition were implemented and tested.
Multi-session training proves to yield higher inter-session validation accuracies than single-session training. Two-posture training proves to be the best postural training (proving the benefit of training on more than one posture), and yields 81.2% inter-posture test accuracy. Five-day training proves to be the best multi-day training, and yields 75.9% inter-day test accuracy. All results are close to the baseline. Moreover, the results of multi-day trainings highlight the phenomenon of user adaptation, indicating that training should also prioritize recent data.
Though not better than the baseline, the achieved classification accuracies rightfully place the 1d-CNN among the candidates for further research.
30
Duggan, Kieran Eamon. "A supervised machine-learning method for detecting steady-state visually evoked potentials for use in brain computer interfaces: A comparative assessment". Master's thesis, University of Cape Town, 2017. http://hdl.handle.net/11427/27335.
Texto completo da fonteResumo:
It is hypothesised that supervised machine learning on the estimated parameters output by a model for visually evoked potentials (VEPs), created by Kremlácek et al. (2002), could be used to classify steady-state visually evoked potentials (SSVEP) by frequency of stimulation. Classification of SSVEPs by stimulus frequency has application in SSVEP-based brain computer interfaces (BCI), where users are presented with flashing stimuli and user intent is decoded by identifying which stimulus the subject is attending to. We investigate the ability of the model of VEPs to fit the initial portions of SSVEPs, which are not yet in a steady state and contain characteristic features of VEPs superimposed with those of a steady state response. In this process the estimated parameters, as a function of the model for a given SSVEP response, were found. These estimated parameters were used to train several support vector machines (SVM) to classify the SSVEPs. Three initialisation conditions for the model are examined for their contribution to the goodness of fit and the subsequent classification accuracy, of the SVMs. It was found that the model was able to fit SSVEPs with a normalised root mean square error (NRMSE) of 27%, this performance did not match the expected NRMSE values of 13% reported by Kremlácek et al. (2002) for fits on VEPs. The fit data was assessed by the machine learning scheme and generated parameters which were classifiable by SVM above a random chance of 14% (Reang 9% to 28%). It was also shown that the selection of initial parameters had no distinct effect on the classification accuracy. Traditional classification approaches using spectral techniques such as Power Spectral Density Analysis (PSDA) and canonical correlation analysis (CCA) require a window period of data above 1 s to perform accurately enough for use in BCIs. The larger the window period of SSVEP data used the more the Information transfer rate (ITR) decreases. Undertaking a successful classification on only the initial 250 ms portions of SSVEP data would lead to an improved ITR and a BCI which is faster to use. Classification of each method was assessed at three SSVEP window periods (0.25, 0.5 and 1 s). Comparison of the three methods revealed that, on a whole CCA outperformed both the PSDA and SVM methods. While PSDA performance was in-line with that of the SVM method. All methods performed poorly at the window period of 0.25 s with an average accuracy converging on random chance - 14%. At the window period of 0.5 s the CCA only marginally outperformed the SVM method and at a time of 1 s the CCA method significantly (p<0.05) outperformed the SVM method. While the SVMs tended to improve with window period the results were not generally significant. It was found that certain SVMs (Representing a unique combination of subject, initial conditions and window period) achieved an accuracy as high as 30%. For a few instances the accuracy was comparable to the CCA method with a significance of 5%. While we were unable to predict which SVM would perform well for a given subject, it was demonstrated that with further refinement this novel method may produce results similar to or better than that of CCA.
31
Österberg, Viktor. "Using Machine Learning to Develop a Quantum-Accurate Inter-Atomic Potential for Large Scale Molecular Dynamics Simulations of Iron under Earth’s Core Conditions". Thesis, KTH, Fysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-298848.
Texto completo da fonteResumo:
Measurements of iron at extreme pressures do not agree on the melting temperature for conditions comparable with those believed to hold at Earth's core. To attempt to determine the stability of relevant lattices, simulations involving a huge amount of particles are needed. In this thesis, a machine learned model is trained to yield results from density functional theory. Different machine learning models are compared. The trained model is then used in molecular dynamics simulations of relevant lattices at a scale too large for density functional theory.Mätningar av järns smälttemperatur under påfrestningar jämförbara med desom tros gälla i jordens kärna överensstämmer ej. För att försöka bestämma stabiliteten av relevanta gitter krävs simulationer av enorma mängder partiklar. I denna tes tränas en maskininlärd modell att återge resultat från Täthetsfunktionalteori. Olika maskininlärningsmodeller jämförs. Den tränade modellen används sedan i Molekyldynamik-simulationer av relevanta gitter som är förstora för Täthetsfunktionalteori.
32
Elhashmi, Rodwan. "Comprehensive Study Toward Energy Opportunity for Buildings Considering Potentials for Using Geothermal and Predicting Chiller Demand". University of Dayton / OhioLINK, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=dayton1589332482268134.
Texto completo da fonte
33
Botros, Andrew Computer Science & Engineering Faculty of Engineering UNSW. "The application of machine intelligence to cochlear implant fitting and the analysis of the auditory nerve response". Awarded By:University of New South Wales. Computer Science & Engineering, 2010. http://handle.unsw.edu.au/1959.4/44707.
Texto completo da fonteResumo:
Effective cochlear implant fitting (or programming) is essential for providing good hearing outcomes, yet it is a subjective and error-prone task. The initial objective of this research was to automate the procedure using the auditory nerve electrically evoked compound action potential (the ECAP) and machine intelligence. The Nucleus?? cochlear implant measures the ECAP via its Neural Response Telemetry (NRT) system. AutoNRT, a commercial intelligent system that measures ECAP thresholds with the Nucleus Freedom implant, was firstly developed in this research. AutoNRT uses decision tree expert systems that automatically recognise ECAPs. The algorithm approaches threshold from lower stimulus levels, ensuring recipient safety during postoperative measurements. Clinical studies have demonstrated success on approximately 95% of electrodes, measured with the same efficacy as a human expert. NRT features other than ECAP threshold, such as the ECAP recovery function, could not be measured with similar success rates, precluding further automation and loudness prediction from data mining results. Despite this outcome, a better application of the ECAP threshold profile towards fitting was established. Since C-level profiles (the contour of maximum acceptable stimulus levels across the implant array) were observed to be flatter than T-level profiles (the contour of minimum audibility), a flattening of the ECAP threshold profile was adopted when applied as a fitting profile at higher stimulus levels. Clinical benefits of this profile scaling technique were demonstrated in a 42 subject study. Data mining results also provided an insight into the ECAP recovery function and refractoriness. It is argued that the ECAP recovery function is heavily influenced by the size of the recruited neural population, with evidence gathered from a computational model of the cat auditory nerve and NRT measurements with 21 human subjects. Slower ECAP recovery, at equal loudness, is a consequence of greater neural recruitment leading to lower mean spike probabilities. This view can explain the counterintuitive association between slower ECAP recovery and greater temporal responsiveness to increasing stimulation rate. This thesis presents the first attempt at achieving completely automated cochlear implant fitting via machine intelligence; a future generation implant, capable of high fidelity auditory system measurements, may realise the ultimate objective.
34
Raphel, Fabien. "Mathematical modelling and learning of biomedical signals for safety pharmacology". Thesis, Sorbonne université, 2022. http://www.theses.fr/2022SORUS116.
Texto completo da fonteResumo:
En tant que branche de la pharmacologie, la pharmacologie de sécurité cardiaque vise à étudier les effets secondaires des composés sur le système cardiaque, à des doses thérapeutiques. Ces études, réalisées par le biais d’expériences in silico, in vitro et in vivo, permettent de sélectionner/rejeter un composé à chaque étape du processus de développement du médicament. Un vaste sous-domaine de la pharmacologie de sécurité cardiaque est consacré à l’étude de l’activité électrique des cellules cardiaques à partir d’expériences in silico et in vitro. Cette activité électrique est la conséquence d’échanges de structures polarisées (principalement des ions) entre le milieu extracellulaire et intracellulaire. Une modification des échanges ioniques induit des changements dans l’activité électrique de la cellule cardiaque qui peuvent être pathologiques (par ex. en générant des arythmies). Une bonne connaissance de ces signaux électriques est donc essentielle pour prévenir les risques d’évènements létaux. Les techniques de patch-clamp sont les méthodes les plus courantes pour enregistrer l’activité électrique d’une cellule cardiaque. Bien que ces signaux électriques soient bien connus, ils sont lents et fastidieux à réaliser, et donc, coûteux. Une alternative récente consiste à considérer les dispositifs de réseaux de microélectrodes (MEA). Développés à l’origine pour l’étude des neurones, leur extension aux cellules cardiaques permet un criblage à haut débit qui n’était pas possible avec les techniques de patch-clamp. Un MEA est une plaque avec des puits dans lesquels des cellules cardiaques (formant un tissu) recouvrent des électrodes. Par conséquent, l’extension de ces dispositifs aux cellules cardiaques permet d’enregistrer l’activité électrique des cellules au niveau du tissu (avant et après l’ajout d’un composé dans les puits). Comme il s’agit d’un nouveau signal, de nombreuses études doivent être menées pour comprendre comment les échanges ioniques induisent cette activité électrique enregistrée, et, enfin, pour procéder à la sélection/rejet d’un composé. Bien que ces signaux soient encore mal connus, des études récentes ont montré des résultats prometteurs dans la prise en compte des MEA dans la pharmacologie de sécurité cardiaque. L’automatisation de la sélection/rejet d’un composé est encore difficile et loin des applications industrielles, ce qui est l’objectif final de ce manuscrit. Mathématiquement, le processus de sélection/rejet peut être considéré comme un problème de classification binaire. Comme dans toute classification supervisée (et dans les tâches d’apprentissage automatique, plus généralement), une entrée doit être définie. Dans notre cas, les séries temporelles des activités électriques cardiaques sont éventuellement longues (minutes ou heures) avec un taux d’échantillonnage élevé (∼ kHz) conduisant à une entrée appartenant à un espace de grande dimension (centaines, milliers ou même plus). De plus, le nombre de données disponibles est encore faible (au plus quelques centaines). Ce régime critique nommé haute dimension/faible taille d’échantillon rend le contexte difficile. Le but de ce manuscrit est de fournir une stratégie systématique pour sélectionner/rejeter des composés d’une manière automatisée, sous les contraintes suivantes: • Traiter le régime de haute dimension/faible taille d’échantillon. • Aucune hypothèse sur la distribution des données. • Exploiter les modèles in silico pour améliorer les performances de classification. • Pas ou peu de paramètres à régler. La première partie du manuscrit est consacrée au contexte, suivie de la description des techniques de patch-clamp et de MEA. Enfin, une description des modèles de potentiel d’action et de potentiel de champ pour réaliser des expériences in silico est donnée. Dans une seconde partie, deux aspects méthodologiques sont développés en respectant au mieux les contraintes définies par le contexte industriel. Le premier décrit une stratégie de [...]As a branch of pharmacology, cardiac safety pharmacology aims at investigating compound side effects on the cardiac system at therapeutic doses. These investigations, made through in silico, in vitro and in vivo experiments, allow to select/reject a compound at each step of the drug development process. A large subdomain of cardiac safety pharmacology is devoted to the study of the electrical activity of cardiac cells based on in silico and in vitro assays. This electrical activity is the consequence of polarised structure exchanges (mainly ions) between the extracellular and intracellular medium. A modification of the ionic exchanges induces changes in the electrical activity of the cardiac cell which can be pathological (e.g. by generating arrhythmia). Strong knowledges of these electrical signals are therefore essential to prevent risk of lethal events. Patch-clamp techniques are the most common methods to record the electrical activity of a cardiac cell. Although these electrical signals are well known, they are slow and tedious to perform, and therefore, expansive. A recent alternative is to consider microelectrode array (MEA) devices. Originally developped for neurons studies, its extension to cardiac cells allows a high throughput screening which was not possible with patch-clamp techniques. It consists of a plate with wells in which cardiac cells (forming a tissue) cover some electrodes. Therefore, the extension of these devices to cardiac cells allow to record the electrical activity of the cells at a tissue level (before and after compound addition into the wells). As a new signal, many studies have to be done to understand how ionic exchanges induce this recorded electrical activity, and, finally, to proceed the selection/rejection of a compound. Despite these signals are still not well known, recent studies have shown promising results in the consideration of MEA into cardiac safety pharmacology. The automation of the compound selection/rejection is still challenging and far from industrial applications, which is the final goal of this manuscript. Mathematically, the selection/rejection process can be seen as a binary classification problem. As in any supervised classification (and machine learning tasks, more generally), an input has to be defined. In our case, time series of the cardiac electrical activities are possibly long (minutes or hours) with a high sampling rate (∼ kHz) leading to an input living in a high-dimensional space (hundreds, thousands or even more). Moreover the number of available data is still low (at most hundreds). This critical regime named high dimension/low sample size make the context challenging. The aim of this manuscript is to provide a systematic strategy to select/reject compounds in an automated way, under the following constraints:• Deal with high dimension/low sample size regime. • No assumptions on the data distributions. • Exploit in silico models to improve the classification performances. • No or few parameters to tune. The first part of the manuscript is devoted to the context, followed by the description of the patch-clamp and MEA technologies. This part ends by the description of action potential and field potential models to perform in silico experiments. In a second part, two methodological aspects are developped, trying to comply, at best, with the constraints of the industrial application. The first one describes a double greedy goal-oriented strategy to reduce the input space based on a score function related to the classification success rate. Comparisons with classical dimension reduction methods such as PCA and PLS (with default parameters) are performed, showing that the proposed method led to better results. The second method consists in the construction of an augmented training set based on a reservoir of simulations, by considering the Hausdorff distance between sets and the maximisation of same score function as in the first method. The proposed strategy [...]
35
Dragoni, Laurent. "Tri de potentiels d'action sur des données neurophysiologiques massives : stratégie d’ensemble actif par fenêtre glissante pour l’estimation de modèles convolutionnels en grande dimension". Thesis, Université Côte d'Azur, 2022. http://www.theses.fr/2022COAZ4016.
Texto completo da fonteResumo:
Au sein du système nerveux, des cellules appelées neurones sont spécialisées dans la communication de l'information. À travers l'émission et la propagation de courants électriques nommés potentiels d'action, les neurones peuvent transmettre l'information dans le corps. Étant donné le rôle prééminent des neurones, afin de mieux comprendre le fonctionnement du système nerveux, une vaste gamme de méthodes ont été proposées pour l'étude de ces cellules. Dans cette thèse, nous nous intéressons à l'analyse de signaux ayant été enregistrés par des électrodes, et plus spécifiquement, des tétrodes et des multi-electrode arrays (MEA). Ces appareils mesurant en général l'activité d'un ensemble de neurones, les signaux enregistrés forment souvent un mélange de l'activité de plusieurs neurones. Afin de gagner plus d'information sur ce type de données, un pré-traitement crucial appelé tri de potentiels d'action est requis pour séparer l'activité de chaque neurone. Actuellement, la procédure générale de tri de potentiels d'action repose sur une procédure en trois étapes : seuillage, extraction de caractéristiques et partitionnement de données. Malheureusement cette méthodologie requiert un grand nombre d'opérations manuelles. De plus, elle devient encore plus difficile à mettre en oeuvre sur de grands volumes de données, en particulier pour des enregistrements de MEA qui ont tendance à présenter davantage de synchronisations de neurones. Dans cette thèse, nous présentons une stratégie de tri de potentiels d'action permettant l'analyse de grands volumes de données et qui requiert peu d'opérations manuelles. Cette stratégie utilise un modèle convolutionnel dont le but est de représenter les signaux mesurés en convolutions temporelles entre deux facteurs : les activations de neurones et les formes de potentiels d'action. L'estimation de ces deux facteurs est généralement traitée par optimisation alternée. Étant la tâche la plus difficile, nous nous concentrons ici sur l'estimation des activations, en supposant que les formes de potentiels d'action sont connues. Le célèbre estimateur Lasso présente d'intéressantes propriétés mathématiques pour la résolution d'un tel problème. Néanmoins son calcul demeure difficile sur des problèmes de grande taille. Nous proposons un algorithme basé sur la stratégie d'ensemble actif afin de calculer efficacement le Lasso. Cet algorithme exploite la structure particulière du problème, déduite de propriétés biologiques, en utilisant des fenêtres glissantes temporelles, lui permettant d'être appliqué en grande dimension. De plus, nous adaptons des résultats théoriques sur le Lasso pour montrer que, sous des hypothèses raisonnables, notre estimateur retrouve le support du vrai vecteur d'activation avec grande probabilité. Nous proposons également des modèles pour la distribution spatiale et des temps d'activations des neurones qui nous permettent de quantifier la taille du problème et de déduire la complexité temporelle théorique de notre algorithme. En particulier, nous obtenons une complexité quasi-linéaire par rapport à la taille du signal enregistré. Finalement nous présentons des expériences numériques illustrant à la fois les résultats théoriques et les performances de notre approcheIn the nervous system, cells called neurons are specialized in the communication of information. Through the generation and propagation of electrical currents named action potentials, neurons are able to transmit information in the body. Given the importance of the neurons, in order to better understand the functioning of the nervous system, a wide range of methods have been proposed for studying those cells. In this thesis, we focus on the analysis of signals which have been recorded by electrodes, and more specifically, tetrodes and multi-electrode arrays (MEA). Since those devices usually record the activity of a set of neurons, the recorded signals are often a mixture of the activity of several neurons. In order to gain more knowledge from this type of data, a crucial pre-processing step called spike sorting is required to separate the activity of each neuron. Nowadays, the general procedure for spike sorting consists in a three steps procedure: thresholding, feature extraction and clustering. Unfortunately this methodology requires a large number of manual operations. Moreover, it becomes even more difficult when treating massive volumes of data, especially MEA recordings which also tend to feature more neuronal synchronizations. In this thesis, we present a spike sorting strategy allowing the analysis of large volumes of data and which requires few manual operations. This strategy makes use of a convolutional model which aims at breaking down the recorded signals as temporal convolutions between two factors: neuron activations and action potential shapes. The estimation of these two factors is usually treated through alternative optimization. Being the most difficult task, we only focus here on the estimation of the activations, assuming that the action potential shapes are known. Estimating the activations is traditionally referred to convolutional sparse coding. The well-known Lasso estimator features interesting mathematical properties for the resolution of such problem. However its computation remains challenging on high dimensional problems. We propose an algorithm based of the working set strategy in order to compute efficiently the Lasso. This algorithm takes advantage of the particular structure of the problem, derived from biological properties, by using temporal sliding windows, allowing it to scale in high dimension. Furthermore, we adapt theoretical results about the Lasso to show that, under reasonable assumptions, our estimator recovers the support of the true activation vector with high probability. We also propose models for both the spatial distribution and activation times of the neurons which allow us to quantify the size of our problem and deduce the theoretical complexity of our algorithm. In particular, we obtain a quasi-linear complexity with respect to the size of the recorded signal. Finally we present numerical results illustrating both the theoretical results and the performances of our approach
36
Schmidt, Eric. "Atomistic modelling of precipitation in Ni-base superalloys". Thesis, University of Cambridge, 2019. https://www.repository.cam.ac.uk/handle/1810/275131.
Texto completo da fonteResumo:
The presence of the ordered $\gamma^{\prime}$ phase ($\text{Ni}_{3}\text{Al}$) in Ni-base superalloys is fundamental to the performance of engineering components such as turbine disks and blades which operate at high temperatures and loads. Hence for these alloys it is important to optimize their microstructure and phase composition. This is typically done by varying their chemistry and heat treatment to achieve an appropriate balance between $\gamma^{\prime}$ content and other constituents such as carbides, borides, oxides and topologically close packed phases. In this work we have set out to investigate the onset of $\gamma^{\prime}$ ordering in Ni-Al single crystals and in Ni-Al bicrystals containing coincidence site lattice grain boundaries (GBs) and we do this at high temperatures, which are representative of typical heat treatment schedules including quenching and annealing. For this we use the atomistic simulation methods of molecular dynamics (MD) and density functional theory (DFT). In the first part of this work we develop robust Bayesian classifiers to identify the $\gamma^{\prime}$ phase in large scale simulation boxes at high temperatures around 1500 K. We observe significant \gamma^{\prime} ordering in the simulations in the form of clusters of $\gamma^{\prime}$-like ordered atoms embedded in a $\gamma$ host solid solution and this happens within 100 ns. Single crystals are found to exhibit the expected homogeneous ordering with slight indications of chemical composition change and a positive correlation between the Al concentration and the concentration of $\gamma^{\prime}$ phase. In general, the ordering is found to take place faster in systems with GBs and preferentially adjacent to the GBs. The sole exception to this is the $\Sigma3 \left(111\right)$ tilt GB, which is a coherent twin. An analysis of the ensemble and time lag average displacements of the GBs reveals mostly `anomalous diffusion' behaviour. Increasing the Al content from pure Ni to Ni 20 at.% Al was found to either consistently increase or decrease the mobility of the GB as seen from the changing slope of the time lag displacement average. The movement of the GB can then be characterized as either `super' or `sub-diffusive' and is interpreted in terms of diffusion induced grain boundary migration, which is posited as a possible precursor to the appearance of serrated edge grain boundaries. In the second part of this work we develop a method for the training of empirical interatomic potentials to capture more elements in the alloy system. We focus on the embedded atom method (EAM) and use the Ni-Al system as a test case. Recently, empirical potentials have been developed based on results from DFT which utilize energies and forces, but neglect the electron densities, which are also available. Noting the importance of electron densities, we propose a route to include them into the training of EAM-type potentials via Bayesian linear regression. Electron density models obtained for structures with a range of bonding types are shown to accurately reproduce the electron densities from DFT. Also, the resulting empirical potentials accurately reproduce DFT energies and forces of all the phases considered within the Ni-Al system. Properties not included in the training process, such as stacking fault energies, are sometimes not reproduced with the desired accuracy and the reasons for this are discussed. General regression issues, known to the machine learning community, are identified as the main difficulty facing further development of empirical potentials using this approach.
37
Ali, Mahammed Ali. "Studie av artificiell intelligens för ökad resurseffektivitet inom produktionsplanering : En studie med fokus på hur nuvarande samt potentiella implementeringar av artificiell intelligens inom produktionsplanering kan öka resurseffektiviteten hos ett tillverkande företag". Thesis, KTH, Maskinkonstruktion (Inst.), 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-299735.
Texto completo da fonteResumo:
Industri 4.0 har medfört stora förändringar och med denna våg av förändringar har artificiell intelligens tillkommit. AI är inget nytt och har forskats på utvecklats sedan den första datorn uppfanns. Tanken var då enligt Alan Turing fadern av datalogi att om en maskin inte kan särskiljas från en människa då är det en AI. Sedan dess har vi sett flera AI modeller slå människan i olika fält och sett AI teknologiers förmåga. Att AI ska implementeras inom den mest innovativa branschen var inte långtsökt. Industriell AI är till skillnad från vanliga AI modeller en kontrollerad process som hittills tillämpats inom begränsade områden. Eftersom standardisering och systematisk tillvägagångsätt kan likställas som synonymer till industriella verksamheter. Är det ingen skillnad på processer inom fabriker, och AI teknologier måste anpassas efter dessa processer. Det har under det senaste decenniet globalt investerats i innovation inom industrier. Länder världen över vill att deras industrier med Industri 4.0 hamnar i framkanten. Där Tyskland introducerade Industri 4.0, USA Smart Manufacturing Leadership Coalition, Kina deras plan kallad China 2025 och EU tillkännagett Factories for the future. Som en konsekvens av dessa enorma satsningar har denna studie som mål att se hur AI kan hjälpa tillverkande företag öka resurseffektiviteten inom produktionsplanering. Eftersom forskningsområdet är relativt nytt kommer studien basera resultaten på fallstudier där ABB och Scania intervjuas. Dock behöver detta område mer forskning.The global introduction of Industry 4.0 has brought with it changes within industry. The indirect consequence of Industry 4.0 being artificial intelligence. The idea of AI is as old as the invention of computers with Alan Turing the father of computer science stating the first description of AI. His thought was that if a machine could be mistaken for a human then the machine was intelligent. The thought being that machine never could outperform humans back then. Now in modern times we have witnessed great feats made by intelligent algorithms where they outperform humans in various fields. For AI to be implemented in industry the most innovative buisness it has to adapt to the workings of indutrial processes. Systematic approach and standardization being two values that strongly represents industries. During the last decade global initiative and investment in innovation of industry. Has led to global competitors such as Germany creating Industry 4.0, The United States creating Smart Manufacturing Leadership Coalition, China introducing their plan called China 2025 and EU with Factories for the future. This paper is a reaction of these enormous investments made into Industry 4.0. The objective of this paper is to examine how AI can help manufacturing enterprises increase their resource efficiency within production planning. Since this field of science stillbeing in its infancy this paper will base its result on interviews made with companies as ABB and Scania. However this field needs more work.
38
Naderi, Darbaghshahi Saeid [Verfasser]. "Exploring the potential of machine learning methods and selection signature analyses for the estimation of genomic breeding values, the estimation of SNP effects and the identification of possible candidate genes in dairy cattle / Saeid Naderi Darbaghshahi". Gießen : Universitätsbibliothek, 2019. http://d-nb.info/1177678365/34.
Texto completo da fonte
39
Ngoungue, Langue Cédric Gacial. "Détection, caractéristiques et prédictibilité des évènements à potentiels forts impacts humains sur les villes ouest-africaines : cas des vagues de chaleur". Electronic Thesis or Diss., université Paris-Saclay, 2023. http://www.theses.fr/2023UPASJ021.
Texto completo da fonteResumo:
Les vagues de chaleur constituent une réelle menace pour l'Homme et son environnement. Sous l'effet du changement climatique, les vagues de chaleur deviendront plus fréquentes et intenses. Les conditions climatiques en Afrique de l'ouest rendent la région favorable aux vagues de chaleur. La première partie de ce travail a été consacrée au monitoring des vagues de chaleur dans 15 villes ouest-Africaines situées sur les régions côtière et continentale. Trois sources d'incertitude ont été identifiées dans la détection d'une vague de chaleur : la première est liée aux données de réanalyse, la seconde repose sur le choix du seuil utilisé pour la définition de la vague de chaleur, et la dernière est la méthodologie utilisée. Les vagues de chaleur nocturnes associées à Tw sont plus fréquentes que celles détectées avec AT,T2m,UTCI. Ceci montre que l'humidité joue un rôle important dans l'occurrence des vagues de chaleur nocturnes, augmentant ainsi le nombre d'événements concomitants (jour et nuit consécutivement) sur le nord du Sahel. La variabilité inter-annuelle des vagues de chaleur dans les différentes régions a mis en évidence pour les 3 indicateurs (AT,T2m,Tw) des années particulièrement chaudes avec une fréquence élevée d'événements: 1998, 2005, 2010, 2016, 2019 et 2020, correspondant pour la plupart aux années El Nino. La région GU est plus touchée par les vagues de chaleur au cours de la dernière décennie (2012-2020) que les régions CONT et ATL. Toutefois, les vagues de chaleur les plus persistantes et les plus intenses se produisent dans la région CONT. Un renforcement de la fréquence, de la durée et de l'intensité des vagues de chaleur est observé durant la dernière décennie. Dans la deuxième partie de ce travail, nous nous sommes intéressés à l'aspect prédictibilité des vagues de chaleur. Une première étude de la prédictibilité des vagues de chaleur a été conduite en utilisant les modèles de prévision intra-saisonnière à saisonnière du CEPMMT et UKMO. Les modèles de prévision présentent de meilleures performances par rapport à une climatologie de référence, principalement pour les prévisions à court terme (deux semaines à l'avance) dans les trois régions. Les vagues de chaleur nocturnes sont plus prévisibles que les vagues de chaleur diurnes. D'après les valeurs de FAR obtenues, seulement 15 à 30% des jours de vague de chaleur prédits par les modèles sont effectivement observés dans les réanalyses, respectivement pour les semaines 5 et 2. Le modèle du CEPMMT émet moins de fausses alertes que UKMO pour les prévisions à court terme. Bien que les modèles démontrent des performances en matière de détection des vagues de chaleur par rapport à une climatologie de référence, leur capacité à prédire l'intensité des événements reste faible même pour de courtes échéances. La prédictibilité des vagues de chaleur a été effectuée en utilisant des méthodes d'apprentissage automatique. La méthode BRF présente de meilleures performances par rapport aux deux autres. Le modèle BRF présente de meilleures performances pour la détection des vagues de chaleur par rapport aux modèles de prévision intra-saisonnière dans les trois régions.La prédictibilité des vagues de chaleur par méthode de prédicteurs de grande échelle tels que la dépression thermique Saharienne (SHL) a été abordée en utilisant deux modèles de prévision saisonnière du centre européen et Météo-France. Le but de cette étude est d'évaluer la représentation et la prévisibilité de la SHL à l'échelle saisonnière. Les modèles sont capables de représenter le cycle saisonnier moyen de la SHL et de capturer certaines caractéristiques de sa variabilité inter-annuelle comme la tendance au réchauffement observée durant les années 2010. En utilisant les outils de correction de biais, les résultats mettent en évidence la capacité des modèles à représenter la variabilité intra-saisonnière de la SHL,mais les performances des modèles restent faibles pour une échéance supérieure à un moisHeat waves (HWs) are a real threat to humans and their environment. Due to climate change, heat waves will become more frequent and more intense. Climatic conditions in West Africa make the region more vulnerable to heat waves. West African cities are highly populated centers, and when it comes to the impact of heat waves on human activities, it's important to study these events at these scales. This study aims to monitor heat waves in major West African cities and evaluate their predictability in subseasonal to seasonal forecast models. The first part of this work focuses on monitoring heat waves in fifteen cities over West Africa located in coastal and continental regions. Three sources of uncertainty encountered in the heat wave detection process were identified: the first related to reanalysis data, the second to the choice of threshold used to define a heat wave, and the last to the methodology adopted. The inter-annual variability of heat waves in the different regions highlighted particularly hot years with a high frequency of heat wave events for all the three indicators AT, T2m,Tw: 1998, 2005, 2010, 2016, 2019 and 2020, mostly corresponding to El Nino years. The GU region has been more affected by heat waves over the past decade (2012-2020) than the CONT and ATL regions. However, the most persistent and intense heat waves occurred in the CONT region. An increase in the frequency, duration and intensity of heat waves has been observed over the last decade (2012-2020), probably due to global warming acting on extreme events." In the second part of this study, we focused on the predictability aspect of heat waves. A preliminary study of the predictability of heat waves has been carried out for the period 2001-2020 using subseasonal to seasonal forecast models from the European Centre for Medium-Range Weather Forecasts (ECMWF) and the United Kingdom Meteorological Office (UKMO). The forecast models perform better than a reference climatology, particularly for short-term forecasts (up to two weeks) in all the three regions. Nighttime heatwaves are more predictable than daytime heatwaves. According to the FAR values, only 15 to 30% of the predicted heatwave days by the models are actually observed in the reanalyses, respectively for lead weeks 5 and 2. This suggests that the models overestimate the duration of heat waves compared with ERA5 reanalysis. ECMWF issues fewer false alarms than UKMO for short-term forecasts. Although the models show skills to detect heat waves compared to a reference climatology, their ability to forecast the intensity of events remains weak even for a short lead time. The predictability of heat waves was performed using machine learning methods. The BRF model demonstrated better heat wave detection skills than subseasonal forecast models in all the three regions. The BRF model considerably improves heat wave detection in forecast models, but on the other hand it generates a high rate of false alarms. The predictability of heat waves using large-scale predictors such as the Saharan Heat Low (SHL) was investigated using two seasonal forecast models: the fifth version of the European Center Seasonal Forecast Model "SEAS5" and the seventh version of the Météo-France Seasonal Forecast Model "MF7". The models show skills on the representation of the mean seasonal cycle of the SHL and capture some characteristics of its inter-annual variability, such as the warming trend observed during the 2010s. SEAS5 makes a more realistic representation of the climatic trend of the SHL compared to MF7. Using bias correction techniques, the results highlight the capacity of the models to represent the intra-seasonal variability of the SHL. Bias correction helps to improve the Continuous Ranked Probability Score (CRPS), but the skills of the model remain low for lead times beyond one month
40
Talevi, Luca. "Sviluppo e test di un sistema BCI SSVEP-based". Bachelor's thesis, Alma Mater Studiorum - Università di Bologna, 2016. http://amslaurea.unibo.it/11636/.
Texto completo da fonteResumo:
Una Brain Computer Interface (BCI) è un dispositivo che permette la misura e l’utilizzo di segnali cerebrali al fine di comandare software e/o periferiche di vario tipo, da semplici videogiochi a complesse protesi robotizzate. Tra i segnali attualmente più utilizzati vi sono i Potenziali Evocati Visivi Steady State (SSVEP), variazioni ritmiche di potenziale elettrico registrabili sulla corteccia visiva primaria con un elettroencefalogramma (EEG) non invasivo; essi sono evocabili attraverso una stimolazione luminosa periodica, e sono caratterizzati da una frequenza di oscillazione pari a quella di stimolazione. Avendo un rapporto segnale rumore (SNR) particolarmente favorevole ed una caratteristica facilmente studiabile, gli SSVEP sono alla base delle più veloci ed immediate BCI attualmente disponibili. All’utente vengono proposte una serie di scelte ciascuna associata ad una stimolazione visiva a diversa frequenza, fra le quali la selezionata si ripresenterà nelle caratteristiche del suo tracciato EEG estratto in tempo reale. L’obiettivo della tesi svolta è stato realizzare un sistema integrato, sviluppato in LabView che implementasse il
paradigma BCI SSVEP-based appena descritto, consentendo di: 1. Configurare la generazione di due stimoli luminosi attraverso l’utilizzo di LED esterni; 2. Sincronizzare l’acquisizione del segnale EEG con tale stimolazione; 3. Estrarre features (attributi caratteristici di ciascuna classe) dal suddetto segnale ed utilizzarle per addestrare un classificatore SVM; 4. Utilizzare il classificatore per realizzare un’interfaccia BCI realtime con feedback per l’utente. Il sistema è stato progettato con alcune delle tecniche più avanzate per l’elaborazione spaziale e temporale del segnale ed il suo funzionamento è stato testato su 4 soggetti sani e comparato alle più moderne BCI SSVEP-based confrontabili rinvenute in letteratura.
41
Michielan, Lisa. "Advance Methodologies in Linear and Nonlinear Quantitative Structure-Activity Relationships (QSARs): from Drug Design to In Silico Toxicology Applications". Doctoral thesis, Università degli studi di Padova, 2010. http://hdl.handle.net/11577/3422242.
Texto completo da fonteResumo:
Novel computational strategies are continuously being demanded by the pharmaceutical industry to assist, improve and speed up the drug discovery process. In this scenario chemoinformatics provide reliable mathematical tools to derive quantitative structure-activity relationships (QSARs), able to describe the correlation between molecular descriptors and various experimental profiles of the compounds. In the last years, nonlinear machine learning approaches have demonstrated a noteworthy predictive capability in several QSAR applications, confirming their superiority over the traditional linear methodologies. Particularly the feasibility of the classification approach has been highlighted in solving complex tasks. Moreover, the introduction of the autocorrelation concept in chemistry allows the structural comparison of the molecules
by using a vectorial fixed-length representation to serve as effective molecular descriptor. In the present thesis we have deeply investigated the wide applicability and the potentialities of nonlinear QSAR strategies, especially in combination with autocorrelation molecular electrostatic potential descriptors projected on the molecular surface. Our intent is arranged in six different case studies that focus on crucial problems in pharmacodynamics, pharmacokinetics and toxicity fields.
The first case study considers the estimation of a physicochemical property, the aqueous solvation free energy, that strictly relates to the pharmacokinetic profile and toxicity of chemicals. Our discussion on pharmacodynamics deals with the prediction of potency and selectivity of human adenosine receptor antagonists (hAR). The adenosine receptor family belongs to GPCR (G protein-coupled receptors) family A, including four different subtypes, referred to as A1, A2A, A2B and A3, which are widely distributed in the tissues. They differentiate for both pharmacological profile and effector coupling. Intensive explorative synthesis and pharmacological evaluation are aimed at discovering potent and selective ligands for each adenosine receptor subtype. In the present thesis, we have considered several pyrazolo-triazolo-pyrimidine and xanthine derivatives, studied as promising adenosine receptor antagonists. Then, a second case study focuses on the comparison and the parallel applicability of linear and nonlinear models to predict the binding affinity of human adenosine receptor A2A antagonists and find a consensus in the prediction results. The following studies evaluate the prediction of both selectivity and binding affinity to A2AR and A3R subtypes by combining classification and regression strategies, to finally investigate the full adenosine receptor potency spectrum and human adenosine receptor subtypes selectivity profile by applying a multilabel classification approach. In the field of pharmacokinetics, and more specifically in metabolism prediction, the use of multi- and single-label classification strategies is involved to analyze the isoform specificity of cytochrome P450 substrates. The results lead to the identification of the appropriate methodology to interpret the real metabolism information, characterized by xenobiotics potentially transformed by multiple cytochrome P450 isoforms. As final case study, we present a computational toxicology investigation. The recent regulatory initiatives due to REACH (Registration, Evaluation, Authorization and Restriction of Chemicals) require the ecotoxicological and risk assessment of chemicals for safety. Most of the current evaluation protocols are based on costly animal experiments. So, chemoinformatic tools are heartily recommended to facilitate the toxicity characterization
of chemical substances. We describe a novel integrated strategy to predict the acute aquatic toxicity through the combination of both toxicokinetic and toxicodynamic behaviors of chemicals, by using a machine learning classification method. The goal is to assign chemicals to different levels of acute aquatic toxicity, providing an appropriate answer to the new regulatory requirements. As preliminary validation of our approach, two toxicokinetic and toxicodynamic models have been applied in series to inspect both aquatic toxicity hazard and mode of action of a set of chemical substances with unknown or uncertain toxicodynamic information, assessing the potential ecological risk and the toxic mechanism.Nuove strategie computazionali vengono continuamente richieste dall'industria farmaceutica per assistere, migliorare e velocizzare il processo di scoperta dei farmaci. In questo scenario la chemoinformatica fornisce affidabili strumenti matematici per ottenere relazioni quantitative struttura-attività (QSAR), in grado di descrivere la correlazione tra descrittori molecolari e vari profili sperimentali dei composti. Negli ultimi anni approcci non lineari di machine learning hanno dimostrato una notevole capacità predittiva in diverse applicazioni QSAR, confermando la loro superiorità sulle tradizionali metodologie lineari. E' stata evidenziata particolarmente la praticabilità dell'approccio di classificazione nel risolvere compiti complessi. Inoltre, l'introduzione del concetto di autocorrelazione in chimica permette il confronto strutturale delle molecole attraverso l'uso di una rappresentazione vettoriale di lunghezza fissa che serve da efficace descrittore molecolare. Nella presente tesi abbiamo studiato approfonditamente l'ampia applicabilità e le potenzialità delle strategie QSAR non lineari, soprattutto in combinazione con i descrittori autocorrelati potenziale elettrostatico molecolare proiettato sulla superficie molecolare. Il nostro intento si articola in sei differenti casi studio, che si concentrano su problemi cruciali nei campi della farmacodinamica, farmacocinetica e tossicologia. Il primo caso studio considera la valutazione di una proprietà fisico-chimica, l'energia libera di solvatazione acquosa, che è strettamente connessa con il profilo farmacocinetico e la tossicità dei composti chimici. La nostra discussione in farmacodinamica riguarda la predizione di potenza e selettività di antagonisti del recettore adenosinico umano (hAR). La famiglia del recettore adenosinico appartiene alla famiglia A di GPCR (recettori accoppiati a proteine G), che include quattro diversi sottotipi, cui ci si riferisce come A1, A2A, A2B e A3, ampiamente distribuiti nei tessuti. Si differenziano sia per profilo farmacologico che per effettore cui sono accoppiati. Le intense sintesi esplorativa e valutazione farmacologica hanno lo scopo di scoprire ligandi potenti e selettivi per ogni sottotipo del recettore adenosinico. Nella presente tesi abbiamo considerato diversi derivati pirazolo-triazolo-pirimidinici e xantinici, studiati come promettenti antagonisti del recettore adenosinico. Quindi, un secondo caso studio si focalizza sul confronto e l'applicabilità in parallelo di modelli lineari e non lineari per predire l'affinità di legame di antagonisti del recettore adenosinico A2A umano e trovare un consenso nei risultati di predizione. Gli studi successivi valutano la predizione sia della selettività che dell'affinità di legame ai sottotipi A2AR e A3R combinando strategie di classificazione e regressione, per studiare infine il completo spettro di potenza del recettore adenosinico e il profilo di selettività per i sottotipi hAR mediante l'applicazione di un approccio di classificazione multilabel. Nel campo della farmacocinetica, e più specificamente nella predizione del metabolismo, è coinvolto l'uso di strategie di classificazione multi- e single-label per analizzare la specificità di isoforma di substrati del citocromo P450. I risultati conducono all'identificazione della metodologia appropriata per interpretare la reale informazione sul metabolismo, caratterizzata da xenobiotici potenzialmente trasformati da multiple isoforme del citocromo P450. Come caso studio finale, presentiamo un'indagine in tossicologia computazionale. Le recenti iniziative regolatorie dovute al REACH (Registration, Evaluation, Authorization and Restriction of Chemicals) richiedono l'accertamento ecotossicologico e del rischio dei composti chimici per la sicurezza. La maggiorparte dei correnti protocolli di valutazione è basata su costosi esperimenti animali. Così, gli strumenti chemoinformatici sono caldamente raccomandati per facilitare la caratterizzazione della tossicità di sostanze chimiche. Noi descriviamo una nuova strategia integrata per predire la tossicità acquatica acuta attraverso la combinazione di entrambi i comportamenti tossicocinetico e tossicodinamico dei composti chimici, utilizzando un metodo di classificazione machine learning. L'obbiettivo è assegnare i composti chimici a diversi livelli di tossicità acquatica acuta, fornendo un'appropriata risposta alle nuove esigenze regolatorie. Come validazione preliminare del nostro approccio, due modelli tossicocinetico e tossicodinamico sono stati applicati in serie per esaminare sia il rischio di tossicità acquatica che il modo d'azione di un set di sostanze chimiche con informazione tossicodinamica sconosciuta o incerta, valutandone il potenziale rischio ecologico ed il meccanismo tossico.
42
Hoffmann, Guillaume. "Mise au point de nouveaux descripteurs théoriques pour la réactivité chimique Can molecular and atomic descriptors predict the electrophilicity of Michael acceptors? On the influence of dynamical effects on reactivity descriptors Predicting experimental electrophilicities from quantum and topological descriptors : a machine learning approach Electrophilicity indices and halogen bonds : some new alternatives to the molecular electrostatic potential". Thesis, Normandie, 2020. http://www.theses.fr/2020NORMR042.
Texto completo da fonteResumo:
L’étude des descripteurs de réactivité globaux, locaux et non locaux d'un système réactif est d’une importance capitale afin de comprendre la réactivité de la totalité des processus chimiques lors d’une réaction. Le but de cette thèse a ainsi été de mettre au point de nouveaux descripteurs de réactivités, ainsi que des modèles de prédiction basés sur ces derniers, afin d’étudier la réactivité chimique. Les principale méthodes théoriques employées ont été la Théorie de la fonctionnelle de la densité conceptuelle (CDFT) et la théorie quantique « Atoms in Molecules » (QTAIM) qui sont toutes deux basées sur la densité électronique. Notre domaine d’étude se place principalement dans le cadre de l’échelle expérimentale de Mayr, qui permet par le biais de mesures cinétiques d’effectuer un classement des molécules par ordre de réactivité. Dans un premier temps, de grande avancées ont été réalisées durant cette thèse vis-à-vis de la prédiction théorique de l’électrophilie expérimentale des accepteurs de Michael. Puis dans un second temps, nous nous sommes intéressées à l’application des descripteurs de réactivité sur la liaison chimique, et particulièrement la liaison halogène. Enfin, une partie de synthèse réalisée au cours de cette thèse est présentée, en proposant une nouvelle voie de synthèse des cations iminium vinyloguesThe study of global, local and non-local reactivity descriptors of a reactive system is of paramount importance in order to understand the reactivity of all chemical processes during a reaction. The goal of this thesis was then to develop new reactivity descriptors, as well as prediction models based on them, in order to study chemical reactivity. The main theoretical methods used were the Conceptual Density Functional Theory (CDFT) and Quantum Theory of “Atoms in Molecules” (QTAIM), which are both based on electron density. Our field of study is mainly within the framework of the Mayr experimental scale, which allows, through kinetic measurements a classification of molecules in order of reactivity. In the first part, great advances were made during this thesis with respect to the theoretical prediction of experimental electrophilicity of Michael acceptors. Then in a second step, we looked at the application of reactivity descriptors on the chemical bond, and in particular the halogen bond. Finally, a part of synthesis carried out during the course of this thesis is presented, by proposing a new way of synthesis of vinylogous iminium cations
43
Skyner, Rachael Elaine. "Hydrate crystal structures, radial distribution functions, and computing solubility". Thesis, University of St Andrews, 2017. http://hdl.handle.net/10023/11746.
Texto completo da fonteResumo:
Solubility prediction usually refers to prediction of the intrinsic aqueous solubility, which is the concentration of an unionised molecule in a saturated aqueous solution at thermodynamic equilibrium at a given temperature. Solubility is determined by structural and energetic components emanating from solid-phase structure and packing interactions, solute–solvent interactions, and structural reorganisation in solution. An overview of the most commonly used methods for solubility prediction is given in Chapter 1. In this thesis, we investigate various approaches to solubility prediction and solvation model development, based on informatics and incorporation of empirical and experimental data. These are of a knowledge-based nature, and specifically incorporate information from the Cambridge Structural Database (CSD). A common problem for solubility prediction is the computational cost associated with accurate models. This issue is usually addressed by use of machine learning and regression models, such as the General Solubility Equation (GSE). These types of models are investigated and discussed in Chapter 3, where we evaluate the reliability of the GSE for a set of structures covering a large area of chemical space. We find that molecular descriptors relating to specific atom or functional group counts in the solute molecule almost always appear in improved regression models. In accordance with the findings of Chapter 3, in Chapter 4 we investigate whether radial distribution functions (RDFs) calculated for atoms (defined according to their immediate chemical environment) with water from organic hydrate crystal structures may give a good indication of interactions applicable to the solution phase, and justify this by comparison of our own RDFs to neutron diffraction data for water and ice. We then apply our RDFs to the theory of the Reference Interaction Site Model (RISM) in Chapter 5, and produce novel models for the calculation of Hydration Free Energies (HFEs).
44
Oliver, Gelabert Antoni. "Desarrollo y aceleración hardware de metodologías de descripción y comparación de compuestos orgánicos". Doctoral thesis, Universitat de les Illes Balears, 2018. http://hdl.handle.net/10803/462902.
Texto completo da fonteResumo:
Introducción
El acelerado ritmo al que se genera y crece la información en la sociedad actual y la posible llegada de la tecnología de transistor a sus límites de tamaño exige la puesta en marcha de soluciones para el procesado eficiente de datos en campos específicos de aplicación.
Contenido
Esta tesis doctoral de carácter transdisciplinar a medio camino entre la ingeniería electrónica y la química computacional presenta soluciones optimizadas en hardware y en software para la construcción y el procesado eficiente de bases de datos moleculares.
En primer lugar se propone y se estudia el funcionamiento de bloques digitales que implementan funciones en lógica pulsante estocástica orientadas a tareas de reconocimiento de objetos. Especialmente se proponen y analizan diseños digitales para la construcción de generadores de números aleatorios (RNG) como base de estos sistemas que han sido implementados en dispositivos Field Programable Gate Array (FPGA).
En segundo lugar se propone y se evalúa un conjunto reducido de descriptores moleculares para la caracterización de compuestos orgánicos y la generación de bases de datos moleculares. Estos descriptores recogen información sobre la distribución de la carga molecular en el espacio y la energía electrostática. Las bases de datos generadas con estos descriptores se han procesado utilizando sistemas de computación convencionales en software y mediante sistemas de computación estocástica implementados en hardware mediante el uso de circuitería digital programable.
Finalmente se proponen optimizaciones para la estimación del potencial electrostático molecular (MEP) y para el cálculo de los puntos de interacción molecular derivados (SSIP).
Conclusiones
Por una parte, los resultados obtenidos ponen de manifiesto la importancia de la uniformidad de los RNG en el período de evaluación para poder implementar sistemas de computación estocástica de alta fiabilidad. Además, los RNG propuestos tienen una naturaleza aperiódica que minimiza las posibles correlaciones entre señales, haciendo que sean adecuados para la implementación de sistemas de computación estocástica.
Por otra parte, el conjunto de descriptores moleculares propuestos PED han demostrado obtener muy buenos resultados en comparación con otros métodos presentes en la literatura. Este hecho se ha discutido mediante los parámetros Area Under The Curve (AUC) y Enrichment Factor (EF) obtenidos de las curvas promedio Receiving Operating Characteristic (ROC). Además, se ha mostrado como la eficacia de los descriptores aumenta cuando se implementan en sistemas de clasificación con aprendizaje supervisado, haciéndolos adecuados para la construcción de un sistema de predicción de dianas terapéuticas eficiente.
En esta tesis, además, se ha determinado que los MEP calculados utilizando la teoría DFT y el conjunto de bases B3LYP/6-31*G en la superficie con densidad electrónica 0,01 au correlacionan bien con datos experimentales debido presumiblemente a la mayor contribución de las propiedades electrostáticas locales reflejadas en el MEP. Las parametrizaciones propuestas en función del tipo de hibridación atómica pueden haber contribuido también a esta mejora. Los cálculos realizados en dichas superficies suponen mejoras en un factor cinco en la velocidad de procesamiento del MEP.
Dado el aceptable ajuste a datos experimentales del método propuesto para el cálculo del MEP aproximado y de los SSIP, éste se puede utilizar con el fin de obtener los SSIP para bases de datos moleculares extensas o en macromoléculas como proteínas de manera muy rápida (ya que la velocidad de procesamiento obtenida puede alcanzar del orden de cinco mil átomos procesados por segundo utilizando un solo procesador).
Estas técnicas resultan de especial interés dadas las numerosas aplicaciones de los SSIP como por ejemplo el cribado virtual de cocristales o la predicción de energías libres en disolución.Introducció El creixement accelerat de les dades en la societat actual i l'arribada de la tecnologia del transistor als límits físics exigeix la proposta de metodologies per al processament eficient de dades. Contingut Aquesta tesi doctoral, de caràcter transdisciplinària i a mig camí entre els camps de l'enginyeria electrònica i la química computacional presenta solucions optimitzades en maquinari i en programari per tal d’accelerar el processament de bases de dades moleculars. En primer lloc es proposa i s'estudia el funcionament de blocs digitals que implementen funcions de lògica polsant estocàstica aplicades a tasques de reconeixement d'objectes. En concret es proposen i analitzen dissenys específics per a la construcció de generadors de nombres aleatoris (RNG) com a sistemes bàsics per al funcionament dels sistemes de computació estocàstics implementats en dispositius programables com les Field Programable Gate Array (FPGA). En segon lloc es proposen i avaluen un conjunt reduït de descriptors moleculars especialment orientats a la caracterització de compostos orgànics. Aquests descriptors reuneixen la informació sobre la distribució de càrrega molecular i les energies electroestàtiques. Les bases de dades generades amb aquests descriptors s’han processat emprant sistemes de computació convencionals en programari i mitjançant sistemes basats en computació estocàstica implementats en maquinari programable. Finalment es proposen optimitzacions per al càlcul del potencial electroestàtic molecular (MEP) calculat mitjançant la teoria del funcional de la densitat (DFT) i dels punts d’interacció que se’n deriven (SSIP). Conclusions Per una banda, els resultats obtinguts posen de manifest la importància de la uniformitat del RNG en el període d’avaluació per a poder implementar sistemes de computació estocàstics d’alta fiabilitat. A més, els RNG proposats presenten una font d’aleatorietat aperiòdica que minimitza les correlacions entre senyals, fent-los adequats per a la implementació de sistemes de computació estocàstica. Per una altra banda, el conjunt de descriptors moleculars proposats PED, han demostrat obtenir molts bons resultats en comparació amb els mètodes presents a la literatura. Aquest fet ha estat discutit mitjançant l’anàlisi dels paràmetres Area Under The Curve (AUC) i Enrichment Factor (EF) de les curves Receiving Operating Characteristic (ROC) analitzades. A més, s’ha mostrat com l’eficàcia dels descriptors augmenta de manera significativa quan s’implementen en sistemes de classificació amb aprenentatge supervisat com les finestres de Parzen, fent-los adequats per a la construcció d’un sistema de predicció de dianes terapèutiques eficient. En aquesta tesi doctoral, a més, s’ha trobat que els MEP calculats mitjançant la teoria DFT i el conjunt de bases B3LYP/6-31*G en la superfície amb densitat electrònica 0,01 au correlacionen bé amb dades experimentals possiblement a causa de la contribució més gran de les propietats electroestàtiques locals reflectides en el MEP. Les parametritzacions proposades en funció del tipus d’hibridació atòmica han contribuït també a la millora dels resultats. Els càlculs realitzats en aquestes superfícies suposen un guany en un factor cinc en la velocitat de processament del MEP. Donat l’acceptable ajust a les dades experimentals del mètode proposat per al càlcul del MEP aproximat i dels SSIP que se’n deriven, aquest procediment es pot emprar per obtenir els SSIP en bases de dades moleculars extenses i en macromolècules (com ara proteïnes) d’una manera molt ràpida (ja que la velocitat de processament obtinguda arriba fins als cinc mil àtoms per segon amb un sol processador). Les tècniques proposades en aquesta tesi doctoral resulten d’interès donades les nombroses aplicacions que tenen els SSIP com per exemple, en el cribratge virtual de cocristalls o en la predicció d’energies lliures en dissolució.
Introduction Because of the generalized data growth in the nowadays digital era and due to the fact that we are possibly living on the last days of the Moore’s law, there exists a good reason for being focused on the development of technical solutions for efficient data processing. Contents In this transdisciplinary thesis between electronic engineering and computational chemistry, it's shown optimal solutions in hardware and software for molecular database processing. On the first hand, there's proposed and studied a set of stochastic computing systems in order to implement ultrafast pattern recognition applications. Specially, it’s proposed and analyzed specific digital designs in order to create digital Random Number Generators (RNG) as a base for stochastic functions. The digital platform used to generate the results is a Field Programmable Gate Array (FPGA). On the second hand, there's proposed and evaluated a set of molecular descriptors in order to create a compact molecular database. The proposed descriptors gather charge and molecular geometry information and they have been used as a database both in software conventional computing and in hardware stochastic computing. Finally, there's a proposed a set of optimizations for Molecular Electrostatic Potential (MEP) and Surface Site Interaction Points (SSIP). Conclusions Firstly, the results show the relevance of the uniformity of the RNG within the evaluation period in order to implement high precision stochastic computing systems. In addition, the proposed RNG have an aperiodic behavior which avoid some potential correlations between stochastic signals. This property makes the proposed RNG suitable for implementation of stochastic computing systems. Secondly, the proposed molecular descriptors PED have demonstrated to provide good results in comparison with other methods that are present in the literature. This has been discussed by the use of Area Under the Curve (AUC) and Enrichment Factor (EF) of averaged Receiving Operating Characteristic (ROC) curves. Furthermore, the performance of the proposed descriptors gets increased when they are implemented in supervised machine learning algorithms making them appropriate for therapeutic target predictions. Thirdly, the efficient molecular database characterization and the usage of stochastic computing circuitry can be used together in order to implement ultrafast information processing systems. On the other hand, in this thesis, it has been found that the MEP calculated by using DFT and B3LYP/6-31*G basis at 0.01 au density surface level has good correlation with experimental data. This fact may be due to the important contribution of local electrostatics and the refinement performed by the parameterization of the MEP as a function of the orbital atom type. Additionally, the proposed calculation over 0.01 au is five times faster than the calculation over 0.002 au. Finally, due to acceptable agreement between experimental data and theoretical results obtained by using the proposed calculation for MEP and SSIP, the proposed method is suitable for being applied in order to quickly process big molecular databases and macromolecules (the processing speed can achieve five thousand molecules per second using a single processor). The proposed techniques have special interest with the purpose of finding the SSIP because the big number of applications they have as for instance in virtual cocrystal screening and calculation of free energies in solution.
45
Zemzemi, Nejib. "Étude théorique et numérique de l'activité électrique du cœur: Applications aux électrocardiogrammes". Phd thesis, Université Paris Sud - Paris XI, 2009. http://tel.archives-ouvertes.fr/tel-00470375.
Texto completo da fonteResumo:
La modélisation du vivant, en particulier la modélisation de l'activité cardiaque, est devenue un défi scientifique majeur. Le but de cette thématique est de mieux comprendre les phénomènes physiologiques et donc d'apporter des solutions à des problèmes cliniques. Nous nous intéressons dans cette thèse à la modélisation et à l'étude numérique de l'activité électrique du cœur, en particulier l'étude des électrocardiogrammes (ECGs). L'onde électrique dans le cœur est gouvernée par un système d'équations de réaction-diffusion appelé modèle bidomaine ce système est couplé à une EDO représentant l'activité cellulaire. Afin simuler des ECGs, nous tenons en compte la propagation de l'onde électrique dans le thorax qui est décrite par une équation de diffusion. Nous commençons par une démonstrer l'existence d'une solution faible du système couplé cœur-thorax pour une classe de modèles ioniques phénoménologiques. Nous prouvons ensuite l'unicité de cette solution sous certaines conditions. Le plus grand apport de cette thèse est l'étude et la simulation numérique du couplage électrique cœur-thorax. Les résultats de simulations sont représentés à l'aide des ECGs. Dans une première partie, nous produisons des simulations pour un cas normal et pour des cas pathologiques (blocs de branche gauche et droit et des arhythmies). Nous étudions également l'impact de certaines hypothèses de modélisation sur les ECGs (couplage faible, utilisation du modèle monodomaine, isotropie, homogénéité cellulaire, comportement résistance-condensateur du péricarde,. . . ). Nous étudions à la fin de cette partie la sensibilité des ECGs par apport aux paramètres du modèle. En deuxième partie, nous effectuons l'analyse numérique de schémas du premier ordre en temps découplant les calculs du potentiel d'action et du potentiel extérieur. Puis, nous combinons ces schémas en temps avec un traîtement explicite du type Robin-Robin des conditions de couplage entre le cœur et le thorax. Nous proposons une analyse de stabilité de ces schémas et nous illustrons les résultats avec des simulations numériques d'ECGs. La dernière partie est consacrée à trois applications. Nous commençons par l'estimation de certains paramètres du modèle (conductivité du thorax et paramètres ioniques). Dans la deuxième application, qui est d'originie industrielle, nous utilisons des méthodes d'apprentissage statistique pour reconstruire des ECGs à partir de mesures ('électrogrammes). Enfin, nous présentons des simulations électro-mécaniques du coeur sur une géométrie réelle dans diverses situations physiologiques et pathologiques. Les indicateurs cliniques, électriques et mécaniques, calculés à partir de ces simulations sont très similaires à ceux observés en réalité.
46
Rothe, Tom. "Machine Learning Potentials - State of the research and potential applications for carbon nanostructures". 2019. https://monarch.qucosa.de/id/qucosa%3A35780.
Texto completo da fonteResumo:
Machine Learning interatomic potentials (ML-IAP) are currently the most promising Non-empirical IAPs for molecular dynamic (MD) simulations. They use Machine Learning (ML) methods to fit the potential energy surface (PES) with large reference datasets of the atomic configurations and their corresponding properties. Promising near quantum mechanical accuracy while being orders of magnitudes faster than first principle methods, ML-IAPs are the new “hot topic” in material
science research.
Unfortunately, most of the available publications require advanced knowledge about ML methods and IAPs, making them hard to understand for beginners and outsiders. This work serves as a plain introduction, providing all the required knowledge about IAPs, ML, and ML-IAPs from the beginning and giving an overview of the most relevant approaches and concepts for building those
potentials. Exemplary a gaussian approximation potential (GAP) for amorphous carbon is used to simulate the defect induced deformation of carbon nanotubes. Comparing the results with published density-functional tight-binding (DFTB) results and own Empirical IAP MD-simulations shows that publicly available ML-IAP can already be used for simulation, being indeed faster than and
nearly as accurate as first-principle methods.
For the future two main challenges appear: First, the availability of ML-IAPs needs to be improved so that they can be easily used in the established MD codes just as the Empirical IAPs. Second, an accurate characterization of the bonds represented in the reference dataset is needed to assure that a potential is suitable for a special application,
otherwise making it a 'black-box' method.:1 Introduction
2 Molecular Dynamics
2.1 Introduction to Molecular Dynamics
2.2 Interatomic Potentials
2.2.1 Development of PES
3 Machine Learning Methods
3.1 Types of Machine Learning
3.2 Building Machine Learning Models
3.2.1 Preprocessing
3.2.2 Learning
3.2.3 Evaluation
3.2.4 Prediction
4 Machine Learning for Molecular Dynamics Simulation
4.1 Definition
4.2 Machine Learning Potentials
4.2.1 Neural Network Potentials
4.2.2 Gaussian Approximation Potential
4.2.3 Spectral Neighbor Analysis Potential
4.2.4 Moment Tensor Potentials
4.3 Comparison of Machine Learning Potentials
4.4 Machine Learning Concepts
4.4.1 On the fly
4.4.2 De novo Exploration
4.4.3 PES-Learn
5 Simulation of defect induced deformation of CNTs
5.1 Methodology
5.2 Results and Discussion
6 Conclusion and Outlook
6.1 Conclusion
6.2 Outlook
47
Turgeon, Stéphanie. "L’analyse appliquée du comportement en autisme et ses enjeux : une évaluation du potentiel de la technologie pour améliorer la pratique et la recherche". Thesis, 2021. http://hdl.handle.net/1866/25604.
Texto completo da fonteResumo:
Le trouble du spectre de l’autisme (TSA) est un trouble neurodéveloppemental caractérisé par des déficits importants de la communication sociale et des interactions sociales ainsi que par la présence de comportements ou d'intérêts restreints et répétitifs. Les données empiriques suggèrent que les interventions découlant de l’analyse appliquée du comportement (AAC) sont les plus efficaces pour intervenir auprès des personnes ayant un TSA. Néanmoins, certaines lacunes en lien avec les interventions découlant de l’analyse du comportement existent. Notamment, le manque d’accessibilité aux services, le manque de connaissances quant aux facteurs sous-jacents à l’efficacité des interventions et les perceptions divergentes de l’AAC freinent son adoption à plus grande échelle. Cette thèse comprend trois études qui mettent à profit la technologie pour mieux comprendre ou améliorer ces enjeux entourant l’AAC.
Dans le cadre ma première étude, les effets d’une formation interactive en ligne qui vise à enseigner aux parents des stratégies découlant de l’AAC pour réduire les comportements problématiques de leur enfant ont été évalués à l’aide d’un devis randomisé contrôlé avec liste d’attente. Les résultats de cette étude soutiennent le potentiel et l’efficacité de la formation pour augmenter la fréquence d’utilisation de stratégies d’intervention découlant de l’AAC par les parents ainsi que pour réduire l’occurrence et la sévérité des comportements problématiques de leur enfant. En revanche, aucune différence significative n’a été observée pour la mesure des pratiques parentales. Certains enjeux éthiques et pratiques entourant la dissémination de la formation en ligne complètement auto-guidées sont discutés.
La deuxième étude de ma thèse doctorale visait donc à montrer comment utiliser des algorithmes d’apprentissage automatique pour identifier les personnes qui sont plus enclines à observer des améliorations suivant une intervention. Plus spécifiquement, l’utilisation de quatre algorithmes d’apprentissage automatique pour prédire les participants ayant pris part à la première étude de cette thèse qui étaient les plus propices à rapporter une diminution des comportements problématiques de leur enfant est démontrée. Cette étude soutient que des algorithmes d’apprentissage automatique peuvent être utilisés avec de petits échantillons pour soutenir la prise de décision des cliniciens et des chercheurs.
La troisième étude cette thèse visait à quantifier l’information sur l’AAC publiée dans quatre sous-forums d’un forum internet, une ressource en ligne souvent utilisée par les familles pour identifier des interventions à utiliser après de leur enfant. Pour atteindre cet objectif, une procédure de forage de données a été réalisée. Les analyses de cette étude appuient que les parents qui fréquentent le forum sont exposés à une proportion importante de messages présentant une désapprobation de l’AAC pour intervenir auprès des personnes ayant un TSA ou bien une description inexacte des principes, méthodes, procédures ou interventions qui en découlent.
Ensemble, les études effectuées dans le cadre de ma thèse doctorale mettent en évidence les bienfaits de la technologie pour l’intervention psychosociale, tant au niveau de l’évaluation que de l’intervention et du transfert de connaissances. Comme souligné dans les trois études de cette thèse, chacun des outils utilisés présente des limites et doit donc être utilisé pour soutenir les cliniciens et les chercheurs, et non pour remplacer leurs interventions et leur jugement clinique. Les études futures doivent continuer à s’intéresser à l’efficacité des outils technologiques, mais également aux facteurs sous-jacents qui favoriseront leur utilisation et aux considérations éthiques liées à leur emploi.Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by significant deficits in social communication and social interactions and by the presence of restricted and repetitive behaviors or interests. Empirical evidence suggests that interventions based on applied behavior analysis (ABA) are the most effective for treating individuals with ASD. Nevertheless, interventions based on behavior analysis present some issues. In particular, intervention services are hard to access, knowledge about the underlying factors of the effectiveness of interventions is lacking and divergent perceptions about of ABA hamper the adoption of the science. This dissertation includes three studies in which technology is used to better understand or improve these issues regarding ABA. As part of my first study, the effects of a fully self-guided interactive web training (IWT) developed for teaching parents of children with ASD ABA-derived strategies to reduce their child's challenging behaviors were evaluated using a randomized waitlist trial. The results of this study support the effectiveness of the IWT for increasing the frequency of parents’ use of behavioral interventions as well as for reducing the frequency and severity of their child’s challenging behaviors. In contrast, no significant difference was observed for the measurement of parenting practices. Ethical and practical consideration regarding the dissemination of fully self-guided online trainings are discussed. The second study of my doctoral thesis aimed to show how to use machine learning algorithms to predict individuals who were most likely to improve following an intervention. Specifically, a demonstration of how to implement four machine learning algorithms to predict the participants from my first study who were the most likely to report a decrease in their child's iv challenging behaviors. This study argues that machine learning algorithms can be used with small samples to support clinicians’ and researchers’ decision making. The third study of my dissertation aimed to quantify the information about ABA published on four subforums of an internet forum; an online resource often used by families to identify potential interventions for their child. This goal was achieved through the use of a data mining procedure. The analyses showed that parents who visited the forum were exposed to a significant proportion of messages that disapproved of ABA for individuals with ASD or that inaccurately described its underlying principles, methods, procedures, or interventions. Together, the studies carried out as part of my doctoral dissertation highlight the benefits of technology to support assessments, interventions, and knowledge gains or transfer within psychosocial practices. As highlighted in the three studies of this dissertation, each of the tools used presents limitations and should therefore be used to support clinicians and researchers, and should not replace their interventions and clinical judgment. Future studies should continue to focus on the effectiveness of technological tools and on the underlying factors that will promote their use. Finally, researchers must reflect on the ethical considerations related to use of technology when working with humans.
48
Nascimento, Matheus Lopes do. "Investigation of Geothermal Potential Zones with Machine Learning in Mainland Portugal". Master's thesis, 2022. http://hdl.handle.net/10362/134617.
Texto completo da fonteResumo:
Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial TechnologiesEarth’s internal heat is explored to produce electricity or used directly in industrial processes or residencies. It is considered to be renewable and cleaner than fossil fuels and has great importance to pursue environmental goals. The exploration phase of geothermal resources is complex and expensive. It requires field surveys, geological, geophysical and geochemical analysis, as well as drilling campaigns. Geospatial data and technologies have been used to target promising sites for further investigations, and helped reduce costs while also pointed to important criteria data related to geothermal potential. Machine learning is a data driven set of technologies that has been successfully used to model environmental parameters, and in the field of geothermal energy it has been used to predict thermal properties of the surface and subsurface. Random Forests and Extreme Gradient Boosting are ensemble machine learning algorithms that have been extensively used in environmental and geological sciences, and have been demonstrated to perform well when predicting thermal properties. This study investigated a methodology that coupled GIS and ML to predict two crucial parameters in geothermal exploration throughout Mainland Portugal: Geothermal gradient and surface Heat flow density. Training data consisted in different types of wells drilled in the study area where the two labels were measured. It was provided by Portugal’s Geology and Energy Laboratory. Features were all publicly available and consisted in geological, hydrogeological, geophysical, weather and terrain data. Data were aggregated in two grids with two spatial resolutions. The results between the two algorithms have been compared and discussed. The most important features that contributed to the models were identified and their relationships with the outputs discussed. The models and the prediction maps over the study area showed the location of zones with higher geothermal gradient and surface heat flow density and can be used to aid geothermal exploration and provide insights for geothermal modelling.
49
Hughes, Zak E., J. C. R. Thacker, A. L. Wilson e P. L. A. Popelier. "Description of Potential Energy Surfaces of Molecules using FFLUX Machine Learning Models". 2018. http://hdl.handle.net/10454/16776.
Texto completo da fonteResumo:
yesA new type of model, FFLUX, to describe the interaction between atoms has been developed as an alternative to traditional force fields. FFLUX models are constructed from applying the kriging machine learning method to the topological energy partitioning method, Interacting Quantum Atoms (IQA). The effect of varying parameters in the construction of the FFLUX models is analyzed, with the most dominant effects found to be the structure of the molecule and the number of conformations used to build the model. Using these models the optimization of a variety of small organic molecules is performed, with sub kJ mol-1 accuracy in the energy of the optimized molecules. The FFLUX models are also evaluated in terms of their performance in describing the potential energy surfaces (PESs) associated with specific degrees of freedoms within molecules. While the accurate description of PESs presents greater challenges than individual minima, FFLUX models are able to achieve errors of <2.5 kJ mol-1 across the full C-C-C-C dihedral PES of n-butane, indicating the future possibilities of the technique.
The full text will be available at the end of the publisher's embargo, 4th Dec 2019.
50
Grey, Talora Bryn. "One step at a time: analysis of neural responses during multi-state tasks". Thesis, 2020. http://hdl.handle.net/1828/11695.
Texto completo da fonteResumo:
Substantial research has been done on the electroencephalogram (EEG) neural signals generated by feedback within a simple choice task, and there is much evidence for the existence of a reward prediction error signal generated in the anterior cingulate cortex of the brain when the outcome of this type of choice does not match expectations. However, less research has been done to date on the neural responses to intermediate outcomes in a multi-step choice task. Here, I investigated the neural signals generated by a complex, non-deterministic task that involved multiple choices before final win/loss feedback in order to see if the observed signals correspond to predictions made by reinforcement learning theory. In Experiment One, I conducted an EEG experiment to record neural signals while participants performed a computerized task designed to elicit the reward positivity, an event-related brain potential (ERP) component thought to be a biological reward prediction error signal. EEG results revealed a difference in amplitude of the reward positivity ERP component between experimental conditions comparing unexpected to expected feedback, as well as an interaction between valence and expectancy of the feedback. Additionally, results of an ERP analysis of the amplitude of the P300 component also showed an interaction between valence and expectancy. In Experiment Two, I used machine learning to classify epoched EEG data from Experiment One into experimental conditions to determine if individual states within the task could be differentiated based solely on the EEG data. My results showed that individual states could be differentiated with above-chance accuracy. I conclude by discussing how these results fit with the predictions made by reinforcement learning theory about the type of task investigated herein, and implications of those findings on our understanding of learning and decision-making in humans.Graduate