Дисертації з теми "Algorithmes de prédiction"
Оформте джерело за APA, MLA, Chicago, Harvard та іншими стилями
Ознайомтеся з топ-50 дисертацій для дослідження на тему "Algorithmes de prédiction".
Біля кожної праці в переліку літератури доступна кнопка «Додати до бібліографії». Скористайтеся нею – і ми автоматично оформимо бібліографічне посилання на обрану працю в потрібному вам стилі цитування: APA, MLA, «Гарвард», «Чикаго», «Ванкувер» тощо.
Також ви можете завантажити повний текст наукової публікації у форматі «.pdf» та прочитати онлайн анотацію до роботи, якщо відповідні параметри наявні в метаданих.
Переглядайте дисертації для різних дисциплін та оформлюйте правильно вашу бібліографію.
Vekemans, Denis. "Algorithmes pour méthodes de prédiction." Lille 1, 1995. http://www.theses.fr/1995LIL10176.
Engelen, Stéfan. "Algorithmes pour la prédiction de structures secondaires d'ARN." Evry-Val d'Essonne, 2006. http://www.theses.fr/2006EVRY0008.
The knowledge of RNA secondary structure is important to understand the relation between structure and function of the RNA. It is made up of a set of helices resulting from the folding of succession of a complementary base pairs. Complexities of existing algorithms is at least of O(n3). This thesis presents an algorithm, called P-DCFold, based on the comparative approach, for the prediction of RNA secondary structures with a complexity of O(n2). In this algorithm, helices are searched recursively using the "divide and conquer" approach. The selection of helices is based on thermodynamic and covariation criteria. The main problem of the comparative approach is the low quality of used alignment. So, P-DCfold use evolutionary models under structure constraints to select correctly aligned sequences. P-DCFold predicts the secondary structure of several RNA with a sensitivity of 0,85 and a sensibility of 0,95
Becquey, Louis. "Algorithmes multi-critères pour la prédiction de structures d'ARN." Electronic Thesis or Diss., université Paris-Saclay, 2021. http://www.theses.fr/2021UPASG065.
Computational RNA structure prediction methods rely on two major algorithmic steps : a sampling step, to propose new structure solutions, and a scoring step to sort the solutions by relevance. A wide diversity of scoring methods exists. Some rely on physical models, some on the similarity to already observed data (so-called data based methods, or knowledge based methods). This thesis proposes structure prediction methods combining two or more scoring criterions, diverse regarding the modelling scale (secondary structure, tertiary structure), their type (theory-based, knowledge-based, compatibility with experimental chemical probing results). The methods describe the Pareto front of the multi-objective optimization problem formed by these criteria. This allows to identify solutions (structures) well scored on each criterion, and to study the correlation between criterions. The presented approaches exploit the latest progresses in the field, like the identification of modules or recurrent interaction networks, and the use of deep learning algorithms. Two neural network architectures (a RNN and a CNN) are adapted from proteins to RNA. A dataset is created to train these networks: RNANet. Two software tools are proposed: the first is called BiORSEO, which predicts the secondary structure based on two criterions (one relative to the structure’s energy, the other relative to the presence of known modules). The second is MOARNA, which predicts coarse-grained 3D structures based on four criterions: energy in 2D and 3D, compatibility with experimental probing results, and with the shape of a known RNA family if one has been identified
Saffarian, Azadeh. "Algorithmes de prédiction et de recherche de multi-structures d'ARN." Phd thesis, Université des Sciences et Technologie de Lille - Lille I, 2011. http://tel.archives-ouvertes.fr/tel-00832700.
Bedrat, Amina. "G4-Hunter : un nouvel algorithme pour la prédiction des G-quadruplexes." Thesis, Bordeaux, 2015. http://www.theses.fr/2015BORD0197/document.
Biologically relevant G4 DNA structures are formed throughout the genome including immunoglobulin switch regions, promoter sequences and telomeric repeats. They can arise when single-stranded G-rich DNA or RNA sequences are exposed during replication, transcription or recombination. Computational analysis using predictive algorithms suggests that the human genome contains approximately 370 000 potential G4-forming sequences. These predictions are generally limited to the standard G3+N(1−7)G3+N(1−7)G3+N(1−7)G3+ description. However, many stable G4s defy this description and escape this consensus; this is the reason why broadening this description should allow the prediction of more G4 loci. We propose an objective score function, G4- hunter, which predicts G4 folding propensity from a linear nucleic acid sequence. The new method focus on guanines clusters and GC asymmetry, taking into account the whole genomic region rather than individual quadruplexes sequences. In parallel with this computational technique, a large scale in vitro experimental work has also been developed to validate the performance of our algorithm in silico on one hundred of different sequences. G4- hunter exhibits unprecedented accuracy and sensitivity and leads us to reevaluate significantly the number of G4-prone sequences in the human genome. G4-hunter also allowed us to predict potential G4 sequences in HIV and Dictyostelium discoideum, which could not be identified by previous computational methods
Bourquard, Thomas. "Exploitation des algorithmes génétiques pour la prédiction de structures protéine-protéine." Paris 11, 2009. http://www.theses.fr/2009PA112302.
Most proteins fulfill their functions through the interaction with one or many partners as nucleic acids, other proteins…. Because most of these interactions are transitory, they are difficult to detect experimentally and obtaining the structure of the complex is generally not possible. Consequently, “in silico prediction” of the existence of these interactions and of the structure of the resulting complex has received a lot of attention in the last decade. However, proteins are very complex objects, and classical computing approaches have lead to computer-time consuming methods, whose accuracy is not sufficient for large scale exploration of the so-called “interactome” of different organisms. In this context development of high-throughput prediction methods for protein-protein docking is needed. We present here the implementation of a new method based on : Two types of formalisms : the Vornonoi and Laguerre tessellations, two simplified geometric models for coarse-grained modeling of complexes. This leads to computation time more reasonable than in atomic representation, the use and optimization of learning algorithms (genetic algorithms) to isolate the most relevant conformation between two two protein parteners, an evaluation method based on clustering of meta-attributes calculated at the interface to sort the best subset of candidate conformations
Voland, Mathieu. "Algorithmes pour la prédiction in silico d'interactions par similarité entre macromolécules biologiques." Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLV014/document.
The action of a drug, or another small biomolecule, is induced by chemical interactions with other macromolecules such as proteins regulating the cell functions. The determination of the set of targets, the macromolecules that could bind the same small molecule, is essential in order to understand molecular mechanisms responsible for the effects of a drug. Indeed, this knowledge could help the drug design process so as to avoid side effects or to find new applications for known drugs. The advances of structural biology provides us with three-dimensional representations of many proteins involved in these interactions, motivating the use of in silico tools to complement or guide further in vitro or in vivo experiments which are both more expansive and time consuming.This research is conducted as part of a collaboration between the DAVID laboratory of the Versailles-Saint-Quentin University, and Bionext SA which offers a software suite to visualize and analyze chemical interactions between biological molecules. The objective is to design an algorithm to predict these interactions for a given compound, using the structures of potential targets. More precisely, starting from a known interaction between a drug and a protein, a new interaction can be inferred with another sufficiently similar protein. This approach consists in the search of a given pattern, the known binding site, across a collection of macromolecules.An algorithm was implemented, BioBind, which rely on a topological representation of the surface of the macromolecules based on the alpha shapes theory. Our surface representation allows to define a concept of region of any shape on the surface. In order to tackle the search of a given pattern region, a heuristic has been developed, consisting in the definition of regular region which is an approximation of a geodesic disk. This circular shape allows for an exhaustive sampling and fast comparison, and any circular region can then be extended to the actual pattern to provide a similarity evaluation with the query binding site.The target prediction problem is formalized as a binary classification problem, where a set of macromolecules is being separated between those predicted to interact and the others, based on their local similarity with the known target. With this point of view, classic metrics can be used to assess performance, and compare our approach with others. Three datasets were used, two of which were extracted from the literature and the other one was designed specifically for our problem emphasizing the pharmacological relevance of the chosen molecules. Our algorithm proves to be more efficient than another state-of-the-art similarity based approach, and our analysis confirms that docking software are not relevant for our target prediction problem when a first target is known, according to our metric
Dieng, Ibnou. "Prédiction de l'interaction génotype x environnement par linéarisation et régression PLS-mixte." Montpellier 2, 2007. http://www.theses.fr/2007MON20019.
Sànchez, Velazquez Jesús Antonio. "Prédiction et évaluation de performance des algorithmes adaptatifs implantés sur machines parallèles." Paris, ENST, 1993. http://www.theses.fr/1993ENST0021.
Suter, Frédéric. "Parallélisme mixte et prédiction de performances sur réseaux hétérogènes de machines parallèles." Lyon, École normale supérieure (sciences), 2002. http://www.theses.fr/2002ENSL0233.
Perriquet, Olivier. "Approche algorithmique pour la prédiction de la structure secondaire des ARN." Lille 1, 2003. https://pepite-depot.univ-lille.fr/LIBRE/Th_Num/2003/50376-2003-211.pdf.
Bourquard, Thomas. "Exploitation des algorithmes génétiques pour la prédiction de structure de complexe protéine-protéine." Phd thesis, Université Paris Sud - Paris XI, 2009. http://tel.archives-ouvertes.fr/tel-00782396.
Mandon, Hugues. "Algorithmes pour la prédiction de stratégies de reprogrammation cellulaire dans les réseaux Booléens." Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLN060.
This thesis explains what is cell reprogramming in Boolean networks, and what are several methods to solve this problem.First, formal definitions of perturbations, perturbation sequences, reprogramming strategies, inevitability and existentiality of the strategies, and of reprogrammability are given, in the scope of Boolean networks.Moreover, a list of actual methods to find cell reprogramming targets is given, both in the scope of Boolean networks and outside of it.Then, it is described how a static analysis of the networks allows for better understanding of their dynamics, and how important strongly connected components of the interaction graph are.From this network with the added information of the attractor list, an algorithm finds a list of variables to perturb, sometimes with the necessity of a precise order.Then, how to construct a new model is explained, allowing to make perturbations sequentially, thus using the Boolean network dynamics between the perturbations.Given the high complexity of this approach, we also explain an in-between approach, where only the attractors of the network can be perturbed, thus allowing for a smaller complexity.Lastly, a case study is done, where biological Boolean networks from literature are used, and on which the different algorithms from the thesis are applied.We show that sequential reprogramming strategies allow for new perturbation sequences, with smaller perturbations than one-step reprogramming strategies
Lafaquière, Vincent. "Compréhension et prédiction de l'énantiosélectivité des lipases." Thesis, Toulouse, INSA, 2010. http://www.theses.fr/2010ISAT0002/document.
This work has been focused on the understanding of the Burkholderia cepacia lipase (BCL) enantioselectivity towards 2-substituted acids which are chiral building blocks of pharmaceutical interest. The main objective of this work was the investigation of the potential role of substrate accessibility toward the buried active site of BCL on enantioselectivity and the development of an engineering procedure for the design of enantioselective mutants. To study further this hypothesis, a novel computational approach, based on motion-planning algorithms, originally used in robotics, was developed. It allows the conformational exploration of constrained high-dimensional spaces and was applied to the computation of trajectories for a set of racemates within the catalytic site. This methodology also enables the identification of residues potentially hindering substrates displacement along the active site. Results obtained in silico were correlated qualitatively with experimental values of enantioselectivity. On the basis of these results, engineering of the narrow active site of BCL has been undertaken to modulate selectively the access of R and S enantiomers to the catalytic triade. An heterologous expression system of BCL in E. coli compatible with production at microplate scale was developed. A library of 57 (3x19) variants targeted at positions Leu17, Val266 and Leu287 was built by iPCR and subsequently screened using a medium-throughput procedure to identify active variants against pNPB hydrolysis. Next, the enantioselectivity of these mutants was evaluated towards a given racemate, the (R,S)-2-chloro ethyl 2-bromophenylacetate, using a novel screening procedure developed in deep wells. Such screening enabled the identification of several variants amongst which the most promising were characterized. Mutants Leu17Ser and Leu17Met showed a remarkable 10-fold increase of their enantioselectivity and a 4- and 5-fold improvement of their specific activity. Compared to the wild-type enzyme, mutant Val266Gly displayed a reversed enantioselectivity for the substrate of interest. Investigation of the trajectories using motion-planning techniques combined to a voxel map representation was carried out. For selected variants, a fair correlation was observed between in silico and experimental results. Moreover, this enabled us to suggest novel combinations of mutations that led to the identification of two double-mutants Leu17Met/Val266Met and Leu17Ser/Leu287Ile showing an enantioselectivity value higher than 150 for the racemic substrate, revealing thus the effiency of the semi-rational strategy
Auger, Nicolas. "Analyse réaliste d'algorithmes standards." Thesis, Paris Est, 2018. http://www.theses.fr/2018PESC1110/document.
At first, we were interested in TimSort, a sorting algorithm which was designed in 2002, at a time where it was hard to imagine new results on sorting. Although it is used in many programming languages, the efficiency of this algorithm has not been studied formally before our work. The fine-grain study of TimSort leads us to take into account, in our theoretical models, some modern features of computer architecture. In particular, we propose a study of the mechanisms of branch prediction. This theoretical analysis allows us to design variants of some elementary algorithms (like binary search or exponentiation by squaring) that rely on this feature to achieve better performance on recent computers. Even if uniform distributions are usually considered for the average case analysis of algorithms, it may not be the best framework for studying sorting algorithms. The choice of using TimSort in many programming languages as Java and Python is probably driven by its efficiency on almost-sorted input. To conclude this dissertation, we propose a mathematical model of non-uniform distribution on permutations, for which permutations that are almost sorted are more likely, and provide a detailed probabilistic analysis
Caignot, Alain. "Prédiction par essais virtuels de l'amortissement dans les structures spatiales." Phd thesis, Cachan, Ecole normale supérieure, 2009. http://www.theses.fr/2009DENS0018.
In the context of a significant cost reduction in the design of space launchers, it is on crucial to control all the factors involved in the dimensionning process. The decrease in mass is compensated by an increase in stiffness and results in a decrease of damping, which is the parameter that determines the level of the dynamic response. At the present time, the damping is taken into account in a global model and most often identified on the final structure. The objective of this work is to improve the launcher design process by introducing the capability to predict damping a priori. In order to do that, the idea is to develop a database containing the dissipation due to the materials and the dissipation relative to the joints in the launcher for each type and each level of solicitation. . . Damping in materials is relatively well-known in the case of the composites which make up the launcher. Therefore, the challenge is the prediction of the damping in the joints where the dissipations can be very important. The experimental approaches are expensive and complex to implement, that is why this work is based on a finite element computation of the joints. This type of simulations is beyond the reach of standard industrial computing codes ans has needed the development of specific parallel computationnal code based on the LATIN method. The robustness of the numerical tool has been studied and its results validated from experimental values obtained in a previous study. Finally, the computation of different joints of the launcher has been done as well as the methodology for integrating these results in the design process of Ariane
Caignot, Alain. "Prédiction par essais virtuels de l'amortissement dans les structures spatiales." Phd thesis, École normale supérieure de Cachan - ENS Cachan, 2009. http://tel.archives-ouvertes.fr/tel-00422291.
Kootbally, Zeïd. "Prédiction des positions de véhicules autonomes dans un environnement routier dynamique." Dijon, 2008. http://www.theses.fr/2008DIJOS064.
The goal of autonomous vehicles research is to build physical systems that can move purposefully and without human intervention in static and dynamic environments, and also in known, partially known and unknown environments. The field of autonomous vehicles is continuing to gain traction both with researchers and practitioners. Funding for research in this area has continued to grow over the past few years, and recent high profile funding opportunities have started to push theoretical research efforts into practical use. Before releasing any autonomous vehicle in the real world, it is important to model the components within a simulated environment and assess the performance of the vehicles in the virtual world. We present the PRIDE framework (Prediction In Dynamic Environments), a hierarchical multiresolutional approach for moving object prediction that incorporates multiple prediction algorithms into a single, unifying framework. PRIDE is based upon the 4D/RCS (Real-time Control System) and provides information to planners at the level of granularity that is appropriate for their planning horizon. The lower levels of the framework utilize estimation theoretic short-term predictions based upon an extended Kalman filter with an associated confidence measure. The upper levels utilize a probabilistic prédiction approach based upon situation recognition with an underlying cost model that provides predictions that incorporate environmental information and constraints. PRIDE is run in the systems' world model independently of the planner and the control system and has knowledge of the road structures via a road network database. The results of the prediction are made available to a planner to allow it to make accurate plans in dynamic environments. We have applied this approach to the visualization tool AutoSim and later on to the MOAST/USARSim framework which incorporates the physics, kinematics and dynamics of vehicles involved in traffic scenarios
Hubans, Christine. "Méthode ab initio de prédiction d'opérons chez les procaryotes et validations biologiques chez les Bordetelles." Lille 1, 2006. https://pepite-depot.univ-lille.fr/RESTREINT/Th_Num/2006/50376_2006_246.pdf.
Hue, Martial. "Méthodes à noyau pour l'annotation automatique et la prédiction d'interaction de structures de protéine." Paris 7, 2011. http://www.theses.fr/2011PA077151.
As large quantities of protein 3D structures are now routinely solved, there is a need for computational tools to automatically annotate protein structures. In this thesis, we investigate several machine learning approaches for this purpose, based on the popular support vector machine (SVM) algorithm. Indeed, the SVM offers several possibilities to overcome the complexity of protein structures, and their interactions. We propose to solve both issues by investigating new positive definite kernels. First, a kernel function for the annotation of protein structures is devised. The kernel is based on a similarity measure called MAMMOTH. Classification tasks corresponding to Enzyme Classification (EC), Structural Classification of Proteins (SCOP), and Gene Ontology (GO) annotation, show that the MAMMOTH kernel significantly outperforms other choices of kernels for protein structures and classifiers. Second, we design a kernel in the context of binary supervised prediction of objects with a specific structure, namely pairs of general objects. The problem of the inference of missing edges in a protein-protein interaction network may be cast in this context. Our results on three benchmarks of interaction between protein structures suggest that the Metric Learning Pairwise Kernel (MLPK), in combination with the MAMMOTH kernel, yield the best performance. Lastly, we introduce a new and efficient learning method for the supervised prediction of protein interaction. A pairwise kernel method is motivated by two previous methods, the Tensor Product Pairwise Kernel (TPPK) and the local model. The connection between the approaches is explicited and the two methods are formulated in a new common framework, that yields to natural generalization by an interpolation
Feng, Lou. "Algorithmes pour l' étude de la structure secondaire des ARN et l'alignement de séquences." Phd thesis, Université Paris Sud - Paris XI, 2012. http://tel.archives-ouvertes.fr/tel-00781416.
Milanese, Jean-Sébastien. "Développement d'un algorithme permettant la prédiction des métastases à partir de mutations germinales et celles du clone fondateur chez des patients atteints du cancer." Master's thesis, Université Laval, 2018. http://hdl.handle.net/20.500.11794/28327.
With the constant progress in neext generation sequencing, the quantity of data available for investigation becomes massive. In parallel, cancer detection methods and treatments remain very specific and barely accurate. Moreover, the patients survival rate are directly linked with tumoral progression and therefore, to cancer detection methods. Despite continual technological advances in recent years, the global cancer mortality rate keeps rising. The creation of new detection methods accessible to all cancer types becomes a necessity. As of now, there is no model available that using sequencing data to predict cancer traits (ex: recurrence, resistance, etc.). The following sections demonstrate the creation of such model using somatic and germline mutations to predict recurrence and its applicability across all cancer types (and even across different diseases). By using gene signatures specific to each cancer types, we were able to obtain an accuracy of 90% (and more) for the cohort where the cancer was recurrent. To our knowledge, this is the first attempt to develop a model that can predict the patient’s prognosis using genome sequencing data. This will affect future studies and improve personalized medicine as well as cancer detection methods.
Djouob, Charles. "Contribution à la synthèse des filtres microondes par une méthode de prédiction fondée sur des données expérimentales : application à la technologie microruban suspendu." Limoges, 1990. http://www.theses.fr/1990LIMO4001.
Dandach, Hoda. "Prédiction de l'espace navigable par l'approche ensembliste pour un véhicule routier." Thesis, Compiègne, 2014. http://www.theses.fr/2014COMP1892/document.
In this thesis, we aim to characterize a vehicle stable state domain, as well as vehicle state estimation, using interval methods.In the first part of this thesis, we are interested in the intelligent vehicle state estimation.The Bayesian approach is one of the most popular and used approaches of estimation. It is based on the calculated probability of the density function which is neither evident nor simple all the time, conditioned on the available measurements.Among the Bayesian approaches, we know the Kalman filter (KF) in its three forms(linear, non linear and unscented). All the Kalman filters assume unimodal Gaussian state and measurement distributions. As an alternative, the Particle Filter(PF) is a sequential Monte Carlo Bayesian estimator. Contrary to Kalman filter,PF is supposed to give more information about the posterior even when it has a multimodal shape or when the noise follows non-Gaussian distribution. However,the PF is very sensitive to the imprecision due by bias or noise, and its efficiency and accuracy depend mainly on the number of propagated particles which can easily and significantly increase as a result of this imprecision. In this part, we introduce the interval framework to deal with the problems of the non-white biased measurements and bounded errors. We use the Box Particle Filter (BPF), an estimator based simultaneously on the interval analysis and on the particle approach. We aim to estimate some immeasurable state from the vehicle dynamics using the bounded error Box Particle algorithm, like the roll angle and the lateral load transfer, which are two dynamic states of the vehicle. BPF gives a guaranteed estimation of the state vector. The box encountering the estimation is guaranteed to encounter thereal value of the estimated variable as well.In the second part of this thesis, we aim to compute a vehicle stable state domain.An algorithm, based on the set inversion principle and the constraints satisfaction,is used. Considering the longitudinal velocity and the side slip angle at the vehicle centre of gravity, we characterize the set of these two state variables that corresponds to a stable behaviour : neither roll-over nor sliding. Concerning the roll-over risk,we use the lateral transfer ratio LTR as a risk indicator. Concerning the sliding risk, we use the wheels side slip angles. All these variables are related geometrically to the longitudinal velocity and the side slip angle at the centre of gravity. Using these constraints, the set inversion principle is applied in order to define the set ofthe state variables where the two mentioned risks are avoided. The algorithm of Sivia is implemented. Knowing the vehicle trajectory, a maximal allowed velocityon every part of this trajectory is deduced
Soucies, Nicolas. "Prédiction de performance d'algorithmes de traitement d'images sur différentes architectures hardwares." Thesis, Paris 6, 2015. http://www.theses.fr/2015PA066129/document.
In computer vision, the choice of a computing architecture is becoming more difficult for image processing experts. Indeed, the number of architectures allowing the computation of image processing algorithms is increasing. Moreover, the number of computer vision applications constrained by computing capacity, power consumption and size is increasing. Furthermore, selecting an hardware architecture, as CPU, GPU or FPGA is also an important issue when considering computer vision applications.The main goal of this study is to predict the system performance in the beginning of a computer vision project. Indeed, for a manufacturer or even a researcher, selecting the computing architecture should be done as soon as possible to minimize the impact on development.A large variety of methods and tools has been developed to predict the performance of computing systems. However, they do not cover a specific area and they cannot predict the performance without analyzing the code or making some benchmarks on architectures. In this works, we specially focus on the prediction of the performance of computer vision algorithms without the need for benchmarking. This allows splitting the image processing algorithms in primitive blocks.In this context, a new paradigm based on splitting every image processing algorithms in primitive blocks has been developed. Furthermore, we propose a method to model the primitive blocks according to the software and hardware parameters. The decomposition in primitive blocks and their modeling was demonstrated to be possible. Herein, the performed experiences, on different architectures, with real data, using algorithms as convolution and wavelets validated the proposed paradigm. This approach is a first step towards the development of a tool allowing to help choosing hardware architecture and optimizing image processing algorithms
Loeffel, Pierre-Xavier. "Algorithmes de machine learning adaptatifs pour flux de données sujets à des changements de concept." Electronic Thesis or Diss., Paris 6, 2017. http://www.theses.fr/2017PA066496.
In this thesis, we investigate the problem of supervised classification on a data stream subject to concept drifts. In order to learn in this environment, we claim that a successful learning algorithm must combine several characteristics. It must be able to learn and adapt continuously, it shouldn’t make any assumption on the nature of the concept or the expected type of drifts and it should be allowed to abstain from prediction when necessary. On-line learning algorithms are the obvious choice to handle data streams. Indeed, their update mechanism allows them to continuously update their learned model by always making use of the latest data. The instance based (IB) structure also has some properties which make it extremely well suited to handle the issue of data streams with drifting concepts. Indeed, IB algorithms make very little assumptions about the nature of the concept they are trying to learn. This grants them a great flexibility which make them likely to be able to learn from a wide range of concepts. Another strength is that storing some of the past observations into memory can bring valuable meta-informations which can be used by an algorithm. Furthermore, the IB structure allows the adaptation process to rely on hard evidences of obsolescence and, by doing so, adaptation to concept changes can happen without the need to explicitly detect the drifts. Finally, in this thesis we stress the importance of allowing the learning algorithm to abstain from prediction in this framework. This is because the drifts can generate a lot of uncertainties and at times, an algorithm might lack the necessary information to accurately predict
Garaud, Jean-Didier. "Développement de méthodes de couplage aéro-thermo-mécanique pour la prédiction d'instabilités dans les structures aérospatiales chaudes." Phd thesis, Université Pierre et Marie Curie - Paris VI, 2008. http://tel.archives-ouvertes.fr/tel-00359175.
Un moteur de couplage est développé pour gérer les aspects logistiques.
Outre l'indépendance spatiale et temporelle des différents codes, il permet de mettre en place rapidement un algorithme de couplage taillé sur mesure pour chaque application.
L'étude d'une tuyère du moteur Vulcain 2, refroidie par écoulement de gaz, sert de fil conducteur applicatif.
Modélisée à haute température par un comportement non-linéaire élasto-visco-plastique, la mécanique couplée est résolue par un algorithme simple.
Au contraire, la thermique se montre problématique, et nécessite l'utilisation conjointe de deux méthodes originales : un pas de temps automatique de couplage, et des conditions de raccord mixtes.
Ces deux cas sont finalement assemblés pour résoudre la question du couplage à trois codes.
Aupy, Guillaume. "Resilient and energy-efficient scheduling algorithms at scale." Thesis, Lyon, École normale supérieure, 2014. http://www.theses.fr/2014ENSL0928.
This thesis deals with two issues for future Exascale platforms, namelyresilience and energy.In the first part of this thesis, we focus on the optimal placement ofperiodic coordinated checkpoints to minimize execution time.We consider fault predictors, a software used by system administratorsthat tries to predict (through the study of passed events) where andwhen faults will strike. In this context, we propose efficientalgorithms, and give a first-order optimal formula for the amount ofwork that should be done between two checkpoints.We then focus on silent data corruption errors. Contrarily to fail-stopfailures, such latent errors cannot be detected immediately, and amechanism to detect them must be provided. We compute the optimal periodin order to minimize the waste.In the second part of the thesis we address the energy consumptionchallenge.The speed scaling technique consists in diminishing the voltage of theprocessor, hence diminishing its execution speed. Unfortunately, it waspointed out that DVFS increases the probability of failures. In thiscontext, we consider the speed scaling technique coupled withreliability-increasing techniques such as re-execution, replication orcheckpointing. For these different problems, we propose variousalgorithms whose efficiency is shown either through thoroughsimulations, or approximation results relatively to the optimalsolution. Finally, we consider the different energetic costs involved inperiodic coordinated checkpointing and compute the optimal period tominimize energy consumption, as we did for execution time
Varet, Suzanne. "Développement de méthodes statistiques pour la prédiction d'un gabarit de signature infrarouge." Phd thesis, Université Paul Sabatier - Toulouse III, 2010. http://tel.archives-ouvertes.fr/tel-00511385.
Loeffel, Pierre-Xavier. "Algorithmes de machine learning adaptatifs pour flux de données sujets à des changements de concept." Thesis, Paris 6, 2017. http://www.theses.fr/2017PA066496/document.
In this thesis, we investigate the problem of supervised classification on a data stream subject to concept drifts. In order to learn in this environment, we claim that a successful learning algorithm must combine several characteristics. It must be able to learn and adapt continuously, it shouldn’t make any assumption on the nature of the concept or the expected type of drifts and it should be allowed to abstain from prediction when necessary. On-line learning algorithms are the obvious choice to handle data streams. Indeed, their update mechanism allows them to continuously update their learned model by always making use of the latest data. The instance based (IB) structure also has some properties which make it extremely well suited to handle the issue of data streams with drifting concepts. Indeed, IB algorithms make very little assumptions about the nature of the concept they are trying to learn. This grants them a great flexibility which make them likely to be able to learn from a wide range of concepts. Another strength is that storing some of the past observations into memory can bring valuable meta-informations which can be used by an algorithm. Furthermore, the IB structure allows the adaptation process to rely on hard evidences of obsolescence and, by doing so, adaptation to concept changes can happen without the need to explicitly detect the drifts. Finally, in this thesis we stress the importance of allowing the learning algorithm to abstain from prediction in this framework. This is because the drifts can generate a lot of uncertainties and at times, an algorithm might lack the necessary information to accurately predict
Alliod, Charlotte. "Conception et modélisation de nouvelles molécules hautement énergétiques en fonction des contraintes réglementaires et environnementales." Thesis, Lyon, 2018. http://www.theses.fr/2018LYSE1035.
For the last two decades, the military research has focused on the improvement of explosive performances, while taking into account their environmental and toxicological impacts. These issues are governed by strict regulations: REACh (Registration, Evaluation, Authorization and Restriction of Chemicals) to ensure a high level of health and environmental protection.Today, it's a major consideration to develop High Energetic Materials (HEM) or molecules who's hazard on human health and environment are reduced. Thus, in collaboration with Airbus Safran Lauchers (ASL), a research program was set up to obtain optimized tools for predicting the potential toxicity of HEM and to design new non-toxic and regulatory molecules.Different in silico methods have been used, including Quantitative Structure Activity Activity Relationships (QSARs) and Machine Learning.The search for structural similarity among molecules is an innovative tool on which we based our predictions in silico. This similarity is obtained thanks to an intelligent algorithm developed within the Pole Rhone Alpin de Bio-Informatique of Lyon which gave rise to a patent. This algorithm allows us to obtain more accurate predictions based on experimental data from European directives
Laroum, Sami. "Prédiction de la localisation des protéines membranaires : méthodes méta-heuristiques pour la détermination du potentiel d'insertion des acides aminés." Phd thesis, Université d'Angers, 2011. http://tel.archives-ouvertes.fr/tel-01064309.
Boulanouar, Ibtissem. "Algorithmes de suivi de cible mobile pour les réseaux de capteurs sans fils." Thesis, Paris Est, 2014. http://www.theses.fr/2014PEST1077/document.
Wireless Sensor Networks (WSN) are a set of tiny autonomous and interconnected devices. These Sensors are scattered in a region of interest to collect information about the surrounding environment depending on the intended application. Nowadays, sensors allow handling more complex data such as multimedia flow. Thus, we observe the emergence of Wireless Multimedia Sensor Networks opening a wider range of applications. In this work, we focus on tracking moving target in these kinds of networks. Target tracking is defined as a two-stage application: detection and localization of the target through its evolution inside an area of interest. This application can be very useful. For example, the presence of an intruder can be detected and its position inside a sensitive area reported, elderly or sick persons carrying sensors can be tracked anytime and so on. Unlike classical monitoring systems, WSN are more flexible and more easy to set up. Moreover, due to their versatility and autonomy they can be used in hostile regions, inaccessible for human. However, these kinds of networks have some limitations: wireless links are not reliable and data processing and transmission are greedy processes in term of energy. To overcome the energy constraint, only the sensors located in target pathway should be activated. Thus, the question is : how to select these sensors to obtain the best compromise between the tracking precision and the energy consumption? This is the question we are trying to answer in this dissertation. Firstly, we focus on communicating targets which have the ability to transmit signals and greatly facilitate the tracking process. The challenge here is to relay the information between the concerned sensors. In order to deal with this challenge, we use a deployment strategy based on virtual forces (VFA: Virtual Forces Algorithm) associated to a distributed tracking algorithm implemented in a cluster-based network. Secondly, we handle a more complex and more frequent case of non-communicating targets. The objective is to detect the presence of such target using movement sensors. We propose the deployment of an heterogeneous wireless sensor networks composed of movement sensors used to detect the target and camera sensors used to locate it. When the target is detected the information is sent to the camera sensors which decide whether to activate or not their cameras based on probabilistic criteria which include the camera orientation angle. Finally, as our last contribution, we specifically focus on target mobility models. These models help us to predict target behaviour and refine the sensor activation process. We use the Extended Kalamn filter as prediction model combined with a change detection mechanism named CuSum (Cumulative Summuray). This mechanism allows to efficiently compute the future target coordinates, and to select which sensors to activate
Laroum, Sami. "Prédiction de la localisation des protéines membranaires : méthodes méta-heuristiques pour la détermination du potentiel d'insertion des acides aminés." Phd thesis, Angers, 2011. https://theses.hal.science/tel-01064309.
In this work, we are interested in the localization of proteins transported towards the endoplasmic reticulum membrane, and more specifically to the recognition of transmembrane segments and signal peptides. By using the last knowledges acquired on the mechanisms of insertion of a segment in the membrane, we propose a discrimination method of these two types of sequences based on the potential of insertion of each amino acid in the membrane. This leads to search for each amino acid a curve giving its potential of insertion according to its place in a window corresponding to the thickness of the membrane. Our goal is to determine "in silico" a curve for each amino acid to obtain the best performances for our method of classification. The optimization, on data sets constructed from data banks of proteins, of the curves is a difficult problem that we address through the meta-heuristic methods. We first present a local search algorithm for learning a set of curves. Its assessment on the different data sets shows good classification results. However, we notice a difficulty in adjusting the curves of certain amino acids. The restriction of the search space with relevant information on amino acids and the introduction of multiple neighborhood allow us to improve the performances of our method and at the same time to stabilize the learnt curves. We also developed a genetic algorithm to explore in a more diversified way the space of search for this problem
Zgheib, Rawad. "Algorithmes adaptatifs d'identification et de reconstruction de processus AR à échantillons manquants." Phd thesis, Université Paris Sud - Paris XI, 2007. http://tel.archives-ouvertes.fr/tel-00273585.
Saule, Cédric. "Modèles combinatoires des structures d'ARN avec ou sans pseudonoeuds, application à la comparaison de structures." Phd thesis, Université Paris Sud - Paris XI, 2011. http://tel.archives-ouvertes.fr/tel-00788467.
Sutton-Charani, Nicolas. "Apprentissage à partir de données et de connaissances incertaines : application à la prédiction de la qualité du caoutchouc." Thesis, Compiègne, 2014. http://www.theses.fr/2014COMP1835/document.
During the learning of predictive models, the quality of available data is essential for the reliability of obtained predictions. These learning data are, in practice very often imperfect or uncertain (imprecise, noised, etc). This PhD thesis is focused on this context where the theory of belief functions is used in order to adapt standard statistical tools to uncertain data.The chosen predictive model is decision trees which are basic classifiers in Artificial Intelligence initially conceived to be built from precise data. The aim of the main methodology developed in this thesis is to generalise decision trees to uncertain data (fuzzy, probabilistic, missing, etc) in input and in output. To realise this extension to uncertain data, the main tool is a likelihood adapted to belief functions,recently presented in the literature, whose behaviour is here studied. The maximisation of this likelihood provide estimators of the trees’ parameters. This maximisation is obtained via the E2M algorithm which is an extension of the EM algorithm to belief functions.The presented methodology, the E2M decision trees, is applied to a real case : the natural rubber quality prediction. The learning data, mainly cultural and climatic,contains many uncertainties which are modelled by belief functions adapted to those imperfections. After a simple descriptiv statistic study of the data, E2M decision trees are built, evaluated and compared to standard decision trees. The taken into account of the data uncertainty slightly improves the predictive accuracy but moreover, the importance of some variables, sparsely studied until now, is highlighted
Brinda, Karel. "Nouvelles techniques informatiques pour la localisation et la classification de données de séquençage haut débit." Thesis, Paris Est, 2016. http://www.theses.fr/2016PESC1027/document.
Since their emergence around 2006, Next-Generation Sequencing technologies have been revolutionizing biological and medical research. Obtaining instantly an extensive amount of short or long reads from almost any biological sample enables detecting genomic variants, revealing the composition of species in a metagenome, deciphering cancer biology, decoding the evolution of living or extinct species, or understanding human migration patterns and human history in general. The pace at which the throughput of sequencing technologies is increasing surpasses the growth of storage and computer capacities, which still creates new computational challenges in NGS data processing. In this thesis, we present novel computational techniques for the problems of read mapping and taxonomic classification. With more than a hundred of published mappers, read mapping might be considered fully solved. However, the vast majority of mappers follow the same paradigm and only little attention has been paid to non-standard mapping approaches. Here, we propound the so-called dynamic mapping that we show to significantly improve the resulting alignments compared to traditional mapping approaches. Dynamic mapping is based on exploiting the information from previously computed alignments, helping to improve the mapping of subsequent reads. We provide the first comprehensive overview of this method and demonstrate its qualities using Dynamic Mapping Simulator, a pipeline that compares various dynamic mapping scenarios to static mapping and iterative referencing. An important component of a dynamic mapper is an online consensus caller, i.e., a program collecting alignment statistics and guiding updates of the reference in the online fashion. We provide OCOCO, the first online consensus caller that implements a smart statistics for individual genomic positions using compact bit counters. Beyond its application to dynamic mapping, OCOCO can be employed as an online SNP caller in various analysis pipelines, enabling calling SNPs from a stream without saving the alignments on disk. Metagenomic classification of NGS reads is another major problem studied in the thesis. Having a database of thousands reference genomes placed on a taxonomic tree, the task is to rapidly assign to tree nodes a huge amount of NGS reads, and possibly estimate the relative abundance of involved species. In this thesis, we propose improved computational techniques for this task. In a series of experiments, we show that spaced seeds consistently improve the classification accuracy. We provide Seed-Kraken, a spaced seed extension of Kraken, the most popular classifier at present. Furthermore, we suggest a new indexing strategy based on a BWT-index, obtaining a much smaller and more informative index compared to Kraken. We provide a modified version of BWA that improves the BWT-index for a quick k-mer look-up
Blot, Guillaume. "Élaboration, parcours et automatisation de traces et savoirs numériques." Thesis, Paris 4, 2017. http://www.theses.fr/2017PA040089.
How access to knowledge can be impacted by Information Technology? In the earlier 2000s, communication tools caused a significant turn : media convergence, participative practices and massive data. In this way, free access to knowledge might tend to be democratized. People seem to regain spaces, reversing traditional top-down model, going from producer to consumer, for the benefit of an horizontal model based on collective intelligence. However, it should not automatically be assumed that this leads to a simple model reversing. Collective intelligence is subject to cognitive biases, leading to potential irrational situations. Formerly, those social mechanisms had limited consequences. Nowadays, digital knowledge are massive communicating spaces, giving birth to new access paths and new cleavages. Why this massive and open knowledge, is actually so selective? I propose to explore this paradox. Massive and constant tracking of traces and individuals hyper-connection, these two facts help organizational structures design, where social dynamics are digitalized in a complex way. These structures formalize human trajectories. On this basis, computer scientists set up prediction algorithms and recommender engines. This way, knowledge access is automatized. It can then be asked about people governance, in this context of infrastructure submission: recording traces, designing knowledge structure and automating algorithms
Travassos-Romano, João Marcos. "Localisation de fréquences bruitées par filtrage adaptatif et implantation d'algorithmes des moindres carrés rapides." Paris 11, 1987. http://www.theses.fr/1987PA112387.
Cénac, Peggy. "Récursivité au carrefour de la modélisation de séquences, des arbres aléatoires, des algorithmes stochastiques et des martingales." Habilitation à diriger des recherches, Université de Bourgogne, 2013. http://tel.archives-ouvertes.fr/tel-00954528.
Samba, Alassane. "Science des données au service des réseaux d'opérateur : proposition de cas d’utilisation, d’outils et de moyens de déploiement." Thesis, Ecole nationale supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, 2018. http://www.theses.fr/2018IMTA0111/document.
The evolution of telecommunications has led today to a proliferation of connected devices and a massification of multimedia services. Faced with this increased demand for service, operators need to adapt the operation of their networks, in order to continue to guarantee a certain level of quality of experience to their users. To do this, operator networks tend towards a more cognitive or autonomic functioning. It is about giving the networks the means to exploit all the information or data at their disposal, helping them to make the best decisions about their services and operations,and even self-manage. It is therefore a questionof introducing artificial intelligence into networks. This requires setting up means to exploit the data, to carry out on them the automatic learning of generalizable models, providing information that can optimize decisions. All these means today constitute a scientific discipline called data science. This thesis fits into a global desire to show the interest of the introduction of data science in different network operating processes. It inlcudes two algorithmic contributions corresponding to use cases of data science for the operator networks, and two software contributions, aiming to facilitate,on the one hand, the analysis, and on the other hand the deployment of the algorithms produced through data science. The conclusive results of these various studies have demonstrated the interest and the feasibility of using data science for the exploitation of operator networks. These results have also been used by related projects
Le, Tan. "Intégration de l'inférence abductive et inductive pour la représentation des connaissances dans les réseaux de gènes." Phd thesis, Toulouse 3, 2014. http://thesesups.ups-tlse.fr/2337/.
Diagnostic reasoning (abductive) and predictive reasoning (inductive) are two methods of reasoning that enable the discovery of new knowledge. When abductive reasoning is the process of finding the best explanation (hypothesis) for a set of observations (Josephson, 1994), the inductive reasoning is the process of predicting, from a set of observations, to find all possible results. These observations may be symptoms of a patient, experiments on genomic and metabolic networks, etc. In this PhD thesis, we are interested in the representation, analysis and synthesis of genomic signaling networks using hypothetical logic. In fact, this thesis focuses on modeling of signaling pathways in response to the DNA double stranded break. To implement the abduction, we use algorithms of production. Then, the default logic is used to build models of minimum representation. These algorithms are proven knowledge discovery on the map of DNA double-strand break. This map is minimal as biological causality graph and allows integrating bio-molecular data
Pochet, Juliette. "Evaluation de performance d’une ligne ferroviaire suburbaine partiellement équipée d’un automatisme CBTC." Thesis, Université Paris-Saclay (ComUE), 2018. http://www.theses.fr/2018SACLC005.
In high-density area, the demand for railway transportation is continuously increasing. Operating companies turn to new intelligent signaling and control systems, such as Communication Based Train Control (CBTC) systems previously deployed on underground systems only. CBTC systems operate trains in automatic pilot and lead to increase the line capacity without expensive modification of infrastructures. They can also include a supervision module in charge of adapting train behavior according to operating objectives and to disturbances, increasing line robustness. In the literature of real-time traffic management, various methods have been proposed to supervise and reschedule trains, on the one hand for underground systems, on the other hand for railway systems. Making the most of the state-of-the-art in both fields, the presented work intend to contribute to the design of supervision and rescheduling functions of CBTC systems operating suburban railway systems. Our approach starts by designing a supervision module for a standard CBTC system. Then, we propose a rescheduling method based on a model predictive control approach and a multi-objective optimization of automatic train commands. In order to evaluate the performances of a railway system, it is necessary to use a microscopic simulation tool including a CBTC model. In this thesis, we present the tool developed by SNCF and named SIMONE. It allows realistic simulation of a railway system and a CBTC system, in terms of functional architecture and dynamics. The presented work has been directly involved in the design and implementation of the tool. Eventually, the proposed rescheduling method was tested with the tool SIMONE on disturbed scenarios. The proposed method was compared to a simple heuristic strategy intending to recover delays. The proposed multi-objective method is able to provide good solutions to the rescheduling problem and over-performs the simple strategy in most cases, with an acceptable process time. We conclude with interesting perspectives for future work
Aouini, Marwen. "Système intelligent utilisant les ondes ultrasonores guidées et le forage de données en vue de la maintenance prédictive." Electronic Thesis or Diss., Université de Lorraine, 2021. http://www.theses.fr/2021LORR0228.
In the Industry 4.0 era, predictive maintenance and internet-of-things are gaining ground. This kind of maintenance does not include yet structural health monitoring (SHM) by guided ultrasonic waves (UGW) in particular. The final objective of the thesis is to develop a tool to enhance this type of maintenance. SHM is an emerging approach that allows continuous monitoring of the structural health of a given structure. It is generally done in three main steps: data acquisition, defect detection and localization (diagnosis) and estimation of the residual life (prognosis). The first step requires the use of non-destructive testing systems such as that of UGW in this thesis. However, these systems were designed to perform spot checks and require the intervention of qualified operators. In this thesis, a system of generation and acquisition of UGW data, allowing among other things to connect the structure to be monitored to a cellular network, has been developed. This allows the construction of databases (which can be heterogeneous) in an automatic and low-cost way. Moreover, a particular attention was paid to the optimization of its power supply to guarantee the most autonomy possible. The second step consists in exploiting these data in order to detect the defect and to localize it. Three approaches have been proposed, depending on the required computing power and the degree of non-stationarity of the data (i.e. due to the instability of the environment of the structure and of the said measurement system). All three approaches are based on the novelty detection technique. In the case where a defect is detected, prediction algorithms of its evolution in time can be used to estimate the residual life of the structure, which is the last monitoring step. Here, a methodology based on a hybrid algorithm, using the empirical mode decomposition technique and an integrated moving average autoregressive model, has been developed. The results obtained on laboratory and in-situ structures show the relevance of the proposed monitoring methodology. Nevertheless, further work is needed to improve the technological maturation of the developed system
Hofleitner, Aude. "Développement d'un modèle d'estimation des variables de trafic urbain basé sur l'utilisation des technologies de géolocalisation." Phd thesis, Université Paris-Est, 2012. http://tel.archives-ouvertes.fr/tel-00798239.
Le, Tan. "Intégration de l'inférence abductive et inductive pour la représentation des connaissances dans les réseaux de gènes." Phd thesis, Université Paul Sabatier - Toulouse III, 2014. http://tel.archives-ouvertes.fr/tel-00996894.
Sahin, Serdar. "Advanced receivers for distributed cooperation in mobile ad hoc networks." Thesis, Toulouse, INPT, 2019. http://www.theses.fr/2019INPT0089.
Mobile ad hoc networks (MANETs) are rapidly deployable wireless communications systems, operating with minimal coordination in order to avoid spectral efficiency losses caused by overhead. Cooperative transmission schemes are attractive for MANETs, but the distributed nature of such protocols comes with an increased level of interference, whose impact is further amplified by the need to push the limits of energy and spectral efficiency. Hence, the impact of interference has to be mitigated through with the use PHY layer signal processing algorithms with reasonable computational complexity. Recent advances in iterative digital receiver design techniques exploit approximate Bayesian inference and derivative message passing techniques to improve the capabilities of well-established turbo detectors. In particular, expectation propagation (EP) is a flexible technique which offers attractive complexity-performance trade-offs in situations where conventional belief propagation is limited by computational complexity. Moreover, thanks to emerging techniques in deep learning, such iterative structures are cast into deep detection networks, where learning the algorithmic hyper-parameters further improves receiver performance. In this thesis, EP-based finite-impulse response decision feedback equalizers are designed, and they achieve significant improvements, especially in high spectral efficiency applications, over more conventional turbo-equalization techniques, while having the advantage of being asymptotically predictable. A framework for designing frequency-domain EP-based receivers is proposed, in order to obtain detection architectures with low computational complexity. This framework is theoretically and numerically analysed with a focus on channel equalization, and then it is also extended to handle detection for time-varying channels and multiple-antenna systems. The design of multiple-user detectors and the impact of channel estimation are also explored to understand the capabilities and limits of this framework. Finally, a finite-length performance prediction method is presented for carrying out link abstraction for the EP-based frequency domain equalizer. The impact of accurate physical layer modelling is evaluated in the context of cooperative broadcasting in tactical MANETs, thanks to a flexible MAC-level simulator
Artero, Sylvaine. "Détection des troubles cognitifs légers (MCI) : algorithmes diagnostiques, dépistage et validité prédictive." Montpellier 1, 2004. http://www.theses.fr/2004MON1T004.
Boumerdassi, Selma. "Mécanismes prédictifs d'allocation de ressources dans les réseaux cellulaires." Versailles-St Quentin en Yvelines, 1998. http://www.theses.fr/1998VERS0020.