Dissertations / Theses on the topic 'Décalage de jeu de données'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 31 dissertations / theses for your research on the topic 'Décalage de jeu de données.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Nodet, Pierre. "Biquality learning : from weakly supervised learning to distribution shifts." Electronic Thesis or Diss., université Paris-Saclay, 2023. http://www.theses.fr/2023UPASG030.
Full textThe field of Learning with weak supervision is called Weakly Supervised Learning and aggregates a variety of situations where the collected ground truth is imperfect. The collected labels may suffer from bad quality, non-adaptability, or insufficient quantity. In this report, we propose a novel taxonomy of Weakly Supervised Learning as a continuous cube called the Weak Supervision Cube that encompasses all of the weaknesses of supervision. To design algorithms capable of handling any weak supervisions, we suppose the availability of a small trusted dataset, without bias and corruption, in addition to the potentially corrupted dataset. The trusted dataset allows the definition of a generic learning framework named Biquality Learning. We review the state-of-the-art of these algorithms that assumed the availability of a small trusted dataset. Under this framework, we propose an algorithm based on Importance Reweighting for Biquality Learning (IRBL). This classifier-agnostic approach is based on the empirical estimation of the Radon-Nikodym derivative (RND), to build a risk-consistent estimator on reweighted untrusted data. Then we extend the proposed framework to dataset shifts. Dataset shifts happen when the data distribution observed at training time is different from what is expected from the data distribution at testing time. So we propose an improved version of IRBL named IRBL2, capable of handling such dataset shifts. Additionally, we propose another algorithm named KPDR based on the same theory but focused on covariate shift instead of the label noise formulation. To diffuse and democratize the Biquality Learning Framework, we release an open-source Python library à la Scikit-Learn for Biquality Learning named biquality-learn
Betta, Mohammed. "Les données évolutives et le décalage : l'analyse statique et l'analyse dynamique." Rennes 2, 1995. http://www.theses.fr/1995REN20010.
Full textThe objective of this thesis is to analyse three index data tables. We are especially interested in the analysis of these data in the case where the third index indicates time. In the first part of this study, we present some of methods used for anlysing three way data. One chapter will be given over to all those that incorporate contiguity structure in their approach. In the second part, we develop a new method of evolutive data analysis under a temporal contiguity constraint. This method is elaborated in two indepedent and complementary steps. First, we introduce the notion of an interval matrix on the left or the right which is used to define a group of similarity indices on individual tables. This is known as static analysis in the second stage, we suggest a new critiria which allows us to determine the subspace where trajectories are represented. This critiria is also defined taking the order structure of time into considerations. We call this dynamic analysis. This thesis concludes by anamysing two examples of evolutive data using two methods, the one we developed and the statis method. A comparison of the obtained results using the two methods is given
Mansiaux, Yohann. "Analyse d'un grand jeu de données en épidémiologie : problématiques et perspectives méthodologiques." Thesis, Paris 6, 2014. http://www.theses.fr/2014PA066272/document.
Full textThe increasing size of datasets is a growing issue in epidemiology. The CoPanFlu-France cohort(1450 subjects), intended to study H1N1 pandemic influenza infection risk as a combination of biolo-gical, environmental, socio-demographic and behavioral factors, and in which hundreds of covariatesare collected for each patient, is a good example. The statistical methods usually employed to exploreassociations have many limits in this context. We compare the contribution of data-driven exploratorymethods, assuming the absence of a priori hypotheses, to hypothesis-driven methods, requiring thedevelopment of preliminary hypotheses.Firstly a data-driven study is presented, assessing the ability to detect influenza infection determi-nants of two data mining methods, the random forests (RF) and the boosted regression trees (BRT), ofthe conventional logistic regression framework (Univariate Followed by Multivariate Logistic Regres-sion - UFMLR) and of the Least Absolute Shrinkage and Selection Operator (LASSO), with penaltyin multivariate logistic regression to achieve a sparse selection of covariates. A simulation approachwas used to estimate the True (TPR) and False (FPR) Positive Rates associated with these methods.Between three and twenty-four determinants of infection were identified, the pre-epidemic antibodytiter being the unique covariate selected with all methods. The mean TPR were the highest for RF(85%) and BRT (80%), followed by the LASSO (up to 78%), while the UFMLR methodology wasinefficient (below 50%). A slight increase of alpha risk (mean FPR up to 9%) was observed for logisticregression-based models, LASSO included, while the mean FPR was 4% for the data-mining methods.Secondly, we propose a hypothesis-driven causal analysis of the infection risk, with a structural-equation model (SEM). We exploited the SEM specificity of modeling latent variables to study verydiverse factors, their relative impact on the infection, as well as their eventual relationships. Only thelatent variables describing host susceptibility (modeled by the pre-epidemic antibody titer) and com-pliance with preventive behaviors were directly associated with infection. The behavioral factors des-cribing risk perception and preventive measures perception positively influenced compliance with pre-ventive behaviors. The intensity (number and duration) of social contacts was not associated with theinfection.This thesis shows the necessity of considering novel statistical approaches for the analysis of largedatasets in epidemiology. Data mining and LASSO are credible alternatives to the tools generally usedto explore associations with a high number of variables. SEM allows the integration of variables des-cribing diverse dimensions and the explicit modeling of their relationships ; these models are thereforeof major interest in a multidisciplinary study as CoPanFlu
Kolmayer, Elisabeth. "Contribution à l'analyse des processus cognitifs mis en jeu dans l'interrogation d'une base de données documentaires." Paris 5, 1997. http://www.theses.fr/1997PA05H051.
Full textDuring the information retrieval process, two topic representations have to interact; one comes from the end-user, the other from the information system. We focus on this interaction. Cognitive psychology gives some tools to analyse the user's domain representation, as research on categorizations and about expert and novice knowledge has shown. An experimental approach with nurses and nursing auxiliaries analyses the effects of two expertise factors: experience and training level. We examine the domain representation in information system through indexing (indexing with descriptors and subject headings). We point out agreement and differences between knowledge organization such as in thesauri, in subject headings indexes and in the user's mind. But by interacting with an actual retrieval device, topic expertise plays a very small role; the important factor is the expertise of the own device. We focus then on the modelling of the information retrieval situation; the problem to face, for the end-user, is not to find the + best match ; between query and index terms, but to develop a representation of his information need convenient with the information system constraints. An information retrieval task is then concieved as a design problem to which the concepts of plans declarative and procedural- can be applied. An experiment tests the efficiency of such a modelling, analysing the seeking process of end-users and librarians interacting with an online catalog. Some ergonomic consequences are considered
Pipien, Sarah. "A la recherche de quasars à grand décalage spectral dans le sondage CFHQSIR." Thesis, Aix-Marseille, 2017. http://www.theses.fr/2017AIXM0376/document.
Full textMy PhD work is focused on the search for high-redshift quasars in the Canada France High-z Quasar Survey in the Near Infrared (CFHQSIR). The main scientific objective of this CFHT Large Program is to search for quasars at redshift z $\sim$ 7 with near-infrared images of the CFHTLS Wide fields acquired with the CFHT WIRCam camera. Firstly, I carried out the photometric calibration of the CFHQSIR images. I performed a detailed analysis of the CFHQSIR data by studying their quality, as well as their noise properties and their depths.Secondly, I computed the number of high-redshift quasars that could be detected with CFHQSIR and the corresponding constraints which could be put on the z $\sim$ 7 quasar luminosity function. Then, I proceeded to the identification of quasar candidates among the many thousands of sources in the 130 square degrees covered by CFHQSIR. Photometric follow-up observations of the candidates revealed about fifty objects, of which fifteen were chosen to be spectroscopically observed with the Very Large Telescope (VLT). Given that this spectroscopic follow-up is not yet completed, the exact nature of these sources will only be known in the coming months. To finish, I applied Bayesian model comparison to my sample in order to complete and consolidate my selection procedure. My candidates were finally classified according to their probability to be a high-redshift quasar. I verified that the majority of the most likely candidates were selected for spectroscopic observations
Bernard, Luc. "Développement d'un jeu de structures de données et de contraintes sémantiques pour la compilation(séparée) du langage ADA." Doctoral thesis, Universite Libre de Bruxelles, 1985. http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/213624.
Full textPonzano, Matteo. "Prévisibilité des épisodes méditerranéens de pluies intenses à l'aide d'un jeu de données de 30 ans de prévisions rétrospectives." Thesis, Toulouse 3, 2019. http://www.theses.fr/2019TOU30329.
Full textThe French Mediterranean region is prone to very intense flash-flood events induced by heavy precipitation events (HPEs), which are responsible for considerable human and material damage. Quantitative precipitation forecasts have improved dramatically in recent years towards quasi-realistic rainfall estimations. Nevertheless, the proper estimation of the uncertainty associated with the physical processes representation remains a challenging issue. In this thesis, we address the predictability of intense precipitation in the French Mediterranean region using a 30-year ensemble hindcast dataset based on the ensemble prediction system PEARP, operational at Météo-France. This reforecast system implements the same model error as PEARP, but initial and boundary conditions are differently assessed. In order to assess the ability of the reforecast to represent the errors of the original model, we first verify this reforecast using some verification scores. The lack of initial condition perturbation makes the ensemble spread of the reforecast lower than the PEARP's one. Though probabilistic forecast scores are weak due to these set-up deficiencies, some skills are observed at 4-days lead time and for very large thresholds. However, the duration of the reforecast dataset and the resolution inherited from the operational model seem to provide enough complexity to the rainfall reforecast distributions. Two post-processing methods, based on quantile mapping and extended logistic regression techniques, are applied to the reforecast. The quantile mapping approach reduces the members biases, but the benefits in terms of probabilistic scores are lower than expected. The calibration procedure using the extended logistic regression approach leads to better probabilistic scores, both for low and large precipitation thresholds. The extended logistic regression fitted on the reforecast as a learning dataset is then applied on the operational ensemble system PEARP over a 4-month period. Though calibrated forecasts skills are not globally improved, some high probability thresholds are slightly improved, suggesting such methodology could be finally efficiently tuned. The last part of this thesis further investigates systematic errors of intense precipitation forecasting using the feature-based metric SAL (Wernli et al. (2008)). This spatial metric applied to the reforecast shows that both amplitude and structure components are controlled by deep convection parametrizations. Indeed, between the two main deep convection schemes implemented in the model, one scheme performs better, in particular for the most extreme events. A remarkable aptitude of the model is emphasised as the ranked distribution of the very intense integrated rainfall features is accurately represented by the model
Charneau, Sébastien. "Approches moléculaires des mécanismes mis en jeu en fin de schizogonie intraérythrocytaire de Plasmodium falciparum (agent du paludisme) par hybridation soustractive suppressive et puces à ADN." Paris, Muséum national d'histoire naturelle, 2005. http://www.theses.fr/2005MNHN0001.
Full textThe aim of this work was to study the differentiation of Plasmodium falciparum merozoite through its transcriptome. First, by suppression subtractive hybridization, we have realised a stage-specific cDNA library. We have identified 40 genes specifically expressed during merozoite morphogenesis, some expected as coding for merozoite proteins (27. 5%), the other coding for proteins of unknown function (72. 5%). We have characterized one of these proteins, the dynamin-2-like. Expressed only during schizogony, it could be a key-molecule of the merozoite differentiation and be involved in mitochondrial division. Second, by microarrays, we have identified 10 genes, as yet unreported, whose expression is significantly modulated by rottlerin, a protein kinase inhibitor that stops merozoite differentiation leading to the parasite lysis
Gogolashvili, Davit. "Global and local Kernel methods for dataset shift, scalable inference and optimization." Electronic Thesis or Diss., Sorbonne université, 2022. https://accesdistant.sorbonne-universite.fr/login?url=https://theses-intra.sorbonne-universite.fr/2022SORUS363v2.pdf.
Full textIn many real world problems, the training data and test data have different distributions. The most common settings for dataset shift often considered in the literature are covariate shift and target shift. In this thesis, we investigate nonparametric models applied to the dataset shift scenario. We develop a novel framework to accelerate Gaussian process regression. In particular, we consider localization kernels at each data point to down-weigh the contributions from other data points that are far away, and we derive the GPR model stemming from the application of such localization operation. We propose a new method for estimating the minimizer and the minimum value of a smooth and strongly convex regression function from the observations contaminated by random noise
Danilo, Céline. "Évaluation des courants de surface océanique au moyen d’un radar à ouverture synthétique." Brest, 2009. http://www.theses.fr/2009BRES2023.
Full textThe aim of this thesis is to extract the radar line-of-sight component of sea surface current from the Doppler shift of a signal received by a synthetic aperture radar on board of ENVISAT. The primary goals of this work are the separation of waves and surface current components to this Doppler shift and, furthermore, the expansion of this analysis to multiple angles of incidence, thereby making it possible to map sea surface currents. This study is limited to vertical polarization and to the radar frequency hand C. Extraction of current signature is based on the assumption that the Doppler shift is caused solely by surface gravity waves and by surface current. Moreover, we initially assume that the current is uniform within the surface of Doppler shift estimation. Under these assumptions, the current component can be deduced from the Doppler shift by substracting the wave component. The latter have relatively short wavelengths, which is typical of wind-sea. Practical use of 10 m wind information for the estimation of waves component is evaluated. A method for extracting surface current information is developed and tested, initially for a constant incidence angle of 23°, on small-sized and isolated ocean images from all over the oceans. A comparison of the monthly average speeds, obtained by on-site current measurements, indicates an RMS error of 0. 35 m/s. As a further step, the same methodology is applied for larger acquisitions (400 km by 400 km minimum) for which the incidence angle ranges from 16 to 42°, This new configuration requires a preliminary analysis of the effect of incidence angle variation on the Doppler shift. We show that an increase of the angle of incidence results in the increase of the relative component of surface current to the Doppler shift, and that the latter also varies according to the direction of wind with respect to the direction of the radar. The estimated surface current component shows a good quantitative correspondence with the same component measured by on-site sensors. This result is encouraging in view of the development of a more systematic technique. Finally, the last part of this work is devoted to a more complex case, namely that of coastal current: here, the hypothesis of a uniform surface current within the estimation surface cannot be maintained anymore due to the great variability of these currents. In this case, we proceed by analyzing maps of Doppler shift of a 1 km resolution. Despite the complexity of this situation, once again, the comparison between Doppler shift-derived velocities and data from tidal current maps indicates a strong potential of SAR images to return high resolution information on surface currents
Karabernou, Si Mahmoud. "Conception et réalisation d'un processeur pour une architecture cellulaire massivement parallèle intégrée." Grenoble INPG, 1993. http://tel.archives-ouvertes.fr/tel-00343216.
Full textSchmoll, Laurence. "Concevoir un scénario de jeu vidéo sérieux pour l'enseignement-apprentissage des langues ou comment dominer un oxymore." Thesis, Strasbourg, 2016. http://www.theses.fr/2016STRAC014.
Full textThis research examines the use of video games in foreign language classrooms. It investigates the integration of this digital medium, originating as an informal leisure activity, in a formal language learning classroom setting. The ‘learning video game’, which refers to both the act of learning and the act of play, can be seen as an oxymoron encompassing seriousness as well as fun. This work focuses on the fun aspect. This work is motivated by the following research question: how can the use of a learning video game foster a playful attitude, hypothetically conducive to a greater involvement of the learner, while also potentially leading to increaseand complexification in student foreign language production and interaction? As a part of the Eveil 3D-project, this study analyses reception from a sample of learners to two learning games which have the same scenario but were completed according to two different game design models. The findings provide both a game design model and a scenario template in order to support foreign language learning through the use of the video game
Bonnot, Titouan. "Réponse du grain de blé à la nutrition azotée et soufrée : étude intégrative des mécanismes moléculaires mis en jeu au cours du développement du grain par des analyses -omiques." Thesis, Clermont-Ferrand 2, 2016. http://www.theses.fr/2016CLF22767/document.
Full textImproving the yield potential of cereals represents a major challenge. In this context, wheat grain quality has to be maintained. Indeed, grain quality is mainly determined by the content and the composition of storage proteins, but there is a strongly negative correlation between yield and grain protein concentration. In addition, grain quality is strongly influenced by the availability of nitrogen and sulfur in soils. Nowadays, the limitation of nitrogen inputs, and also the sulfur deficiency recently observed in soils represent major difficulties to control the quality. Therefore, understanding of molecular mechanisms controlling grain development and accumulation of storage proteins in response to nitrogen and sulfur supply is a major issue. The objective of this thesis was to create knowledge on the comprehension of these regulatory mechanisms. For this purpose, the best strategy to identify molecular actors involved in these processes consisted of -omics approaches. In our studies, the nuclear proteome was an important target. Among these proteins, we revealed some transcriptional regulators likely to be involved in the control of the accumulation of grain storage compounds. Using an approach combining proteomic, transcriptomic and metabolomic data, the characterization of the integrative grain response to the nitrogen and sulfur supply was obtained. Besides, our studies clearly confirmed the major influence of sulfur in the control of the nitrogen/sulfur balance that determines the grain storage protein composition. Among the changes observed in the cell metabolism, some genes were disturbed by the modification of this balance. Thus these genes could coordinate the adjustment of grain composition in response to nutritional deficiencies. These new results contribute in facing the challenge of maintaining wheat grain quality with sustainable agriculture
Machado, Daniel. "Improving automated redshift detection in the low signal-to-noise regime for Large Sky Surveys." Thesis, Paris 11, 2015. http://www.theses.fr/2014PA112427.
Full textSummary: Redshift is the primary measure by which astronomers can map the Universe in the radial direction. In order to test the assumptions of homogeneity and isotropy, accurate redshifts of galaxies are needed, and for a great many of them. Additionally different cosmological models can only be distinguished by careful observations of the large scale structure traced by these galaxies. Large sky surveys are the only mechanism by which redshifts for a large number of galaxies can be obtained. Accurate redshift estimation is additionally required for many other fields of astronomy including but not limited to: weak lensing, studies of dark matter haloes, galaxy morphology studies, chemical evolution studies, photometric calibration, and studies of large scale structure and galaxy clustering.Problems exist in all surveys at the dim limit of observation, which usually corresponds to the higher redshift objects in the survey, where noise becomes problematic. Magnitude or signal-to-noise ratio cuts are often employed in order to eliminate potentially troublesome objects; such a procedure is a blunt tool for separating good redshift candidates from ones likely to be inaccurate.In this thesis we develop an algorithm to tackle redshift estimation of galaxy spectra in the low signal-to-noise regime. The first part of this thesis introduces the concepts of denoising, particularly False Detection Rate denoising, wavelet transforms and redshift estimation algorithms. The second part details how these concepts are united into the Darth Fader (Denoised and Automatic Redshifts THresholded with a FAlse DEtection Rate) algorithm. The final parts of this thesis apply the algorithm both to idealised synthetic data generated from the COSMOS Mock Catalogue, and to a subset of real data from the WiggleZ survey.We show that Darth Fader can operate effectively at low signal-to-noise given an appropriate choice of FDR parameter for denoising, and an appropriate feature-counting criterion. We also show that Darth Fader can remove the continua of spectra effectively at low signal-to-noise for the purposes of redshift estimation by cross-correlation. Additionally we show from tests on spectra from the WiggleZ survey that our algorithm has the ability to process a substantial subset of that data without the need for visual inspection (to which the entire WiggleZ spectral survey has been subjected), and to a high degree of accuracy. We conclude that the Darth Fader algorithm has potential to be used in large-sky survey pipelines, particularly where signal-to-noise is expected to be poor
Awwad, Tarek. "Context-aware worker selection for efficient quality control in crowdsourcing." Thesis, Lyon, 2018. http://www.theses.fr/2018LYSEI099/document.
Full textCrowdsourcing has proved its ability to address large scale data collection tasks at a low cost and in a short time. However, due to the dependence on unknown workers, the quality of the crowdsourcing process is questionable and must be controlled. Indeed, maintaining the efficiency of crowdsourcing requires the time and cost overhead related to this quality control to stay low. Current quality control techniques suffer from high time and budget overheads and from their dependency on prior knowledge about individual workers. In this thesis, we address these limitation by proposing the CAWS (Context-Aware Worker Selection) method which operates in two phases: in an offline phase, the correlations between the worker declarative profiles and the task types are learned. Then, in an online phase, the learned profile models are used to select the most reliable online workers for the incoming tasks depending on their types. Using declarative profiles helps eliminate any probing process, which reduces the time and the budget while maintaining the crowdsourcing quality. In order to evaluate CAWS, we introduce an information-rich dataset called CrowdED (Crowdsourcing Evaluation Dataset). The generation of CrowdED relies on a constrained sampling approach that allows to produce a dataset which respects the requester budget and type constraints. Through its generality and richness, CrowdED helps also in plugging the benchmarking gap present in the crowdsourcing community. Using CrowdED, we evaluate the performance of CAWS in terms of the quality, the time and the budget gain. Results shows that automatic grouping is able to achieve a learning quality similar to job-based grouping, and that CAWS is able to outperform the state-of-the-art profile-based worker selection when it comes to quality, especially when strong budget ant time constraints exist. Finally, we propose CREX (CReate Enrich eXtend) which provides the tools to select and sample input tasks and to automatically generate custom crowdsourcing campaign sites in order to extend and enrich CrowdED
Danilo, Céline. "Évaluation des courants de surface océanique au moyen d'un radar à ouverture synthétique." Phd thesis, Université de Bretagne occidentale - Brest, 2009. http://tel.archives-ouvertes.fr/tel-00560597.
Full textPavaux, Alice. "Inductive, Functional and Non-Linear Types in Ludics." Thesis, Sorbonne Paris Cité, 2017. http://www.theses.fr/2017USPCD092.
Full textThis thesis investigates the types of ludics. Within the context of the Curry–Howard correspondence,l udics is a framework in which the dynamic aspects of both logic and programming can be studied. The basic objects, called designs, are untyped infinitary proofs that can also beseen as strategies from the perspective of game semantics, and a type or behaviour is a set of designs well-behaved with respect to interaction. We are interested in observing the interactive properties of behaviours. Our attention is particularly focused on behaviours representing the types of data and functions, and on non-linear behaviours which allow the duplication of objects. A new internal completeness result for infinite unions unveils the structure of inductive data types. Thanks to an analysis of the visitable paths, i.e., the possible execution traces, we prove that inductive and functional behaviours are regular, paving the way for a characterisation of MALL in ludics. We also show that a functional behaviour is pure, a property ensuring the safety of typing, if and only if it is not a type of functions taking functions as argument. Finally,we set the bases for a precise study of non-linearity in ludics by recovering a form of internal completeness and discussing the visitable paths
Vermeulen, Mathieu. "Une approche meta-design des learning games pour développer leur usage." Electronic Thesis or Diss., Sorbonne université, 2018. http://www.theses.fr/2018SORUS093.
Full textThis thesis in computer science is in the field of Technology Enhanced Learning (TEL) and more specifically in the field of Learning Games (LG), serious games dedicated to learning. It deals with their design, tools and models to facilitate it, and their use. To tackle this problem, we use the meta-design, an approach aiming to strongly involve the end users in the design stage but also in the use stage. To implement this approach with teachers, whom we consider as end users of LG, we propose different iterations of a simple and representable LG model to facilitate the collaborative design of these TELs, but also their reengineering. After a first iteration, the second iteration propose a model named DISC and the associated design method. They were tested in the co-design of a learning game used by teachers of higher education in the context of a MOOC and as an additional activity of a course. To involve teachers in the use stage, we propose to articulate this model with a learners’ traces visualizations tool to detect problematic pattern and, thus, facilitate the reengineering process of LG, the visualizations allowing traces analysis collected during the use stage. To carry out this research work, we chose to work with the THEDRE method which proposes an iterative research cycle supported by the feedback of indicators evaluating the process throughout the method. This continuous improvement, supported by the experiments, allow us to validate our propositions about the meta-design for the learning games
Vaquette, Geoffrey. "Reconnaissance robuste d'activités humaines par vision." Thesis, Sorbonne université, 2018. http://www.theses.fr/2018SORUS090.
Full textThis thesis focuses on supervised activity segmentation from video streams within application context of smart homes. Three semantic levels are defined, namely gesture, action and activity, this thesis focuses mainly on the latter. Based on the Deeply Optimized Hough Transform paridigm, three fusion levels are introduced in order to benefit from various modalities. A review of existing action based datasets is presented and the lack of activity detection oriented database is noticed. Then, a new dataset is introduced. It is composed of unsegmented long time range daily activities and has been recorded in a realistic environment. Finaly, a hierarchical activity detection method is proposed aiming to detect high level activities from unsupervised action detection
Chamekh, Rabeb. "Stratégies de jeux pour quelques problèmes inverses." Thesis, Université Côte d'Azur (ComUE), 2019. http://www.theses.fr/2019AZUR4103.
Full textIn this PHD-Thesis, we focused on solving the coupling problem of data completion and parameter identification. The Cauchy problem is a problem of identification of boundary condition on a part of the boundary from overabundant data on the remaining part. Parameter identification is a problem of the system parameter. These two problems are known to be ill-posed in the sense of Hadamard. This Thesis is divided into four parts. The first part is dedicated to a bibliography study. In the second chapter, we applied the game theory on the resolution of the coupling problem of data completion and the conductivity identification in electrocardiography. We talked about the identifiability of the conductivity. We have shown the uniqueness of this parameter using only the Cauchy data on a part of the edge. Our numerical experiments target medical applications in electrocardiography. We applied our procedure in a two-dimensional and three-dimensional thorax. The third part is dedicated to the resolution of the coupling problem in linear elasticity applying the game theory. A numerical study has been done where we considered a particular configuration to ensure the parameters identifiability. In the last part, we are interested in a problem of thermoelasticity. It’s about coupling two different disciplines : thermal and elasticity. The problem of crack identification is a natural application in this case
Paudel, Subodh. "Methodology to estimate building energy consumption using artificial intelligence." Thesis, Nantes, Ecole des Mines, 2016. http://www.theses.fr/2016EMNA0237/document.
Full textHigh-energy efficiency building standards (as Low energy building LEB) to improve building consumption have drawn significant attention. Building standards is basically focused on improving thermal performance of envelope and high heat capacity thus creating a higher thermal inertia. However, LEB concept introduces alarge time constant as well as large heat capacity resulting in a slower rate of heat transfer between interior of building and outdoor environment. Therefore, it is challenging to estimate and predict thermal energy demand for such LEBs. This work focuses on artificial intelligence (AI) models to predict energy consumptionof LEBs. We consider two kinds of AI modeling approaches: “all data” and “relevant data”. The “all data” uses all available data and “relevant data” uses a small representative day dataset and addresses the complexity of building non-linear dynamics by introducing past day climatic impacts behavior. This extraction is based on either simple physical understanding: Heating Degree Day (HDD), modified HDD or pattern recognition methods: Frechet Distance and Dynamic Time Warping (DTW). Four AI techniques have been considered: Artificial Neural Network (ANN), Support Vector Machine (SVM), Boosted Ensemble Decision Tree (BEDT) and Random forest (RF). In a first part, numerical simulations for six buildings (heat demand in the range [25 – 85 kWh/m².yr]) have been performed. The approach “relevant data” with (DTW, SVM) shows the best results. Real data of the building “Ecole des Mines de Nantes” proves the approach is still relevant
Yang, Wenlu. "Personalized physiological-based emotion recognition and implementation on hardware." Thesis, Sorbonne université, 2018. http://www.theses.fr/2018SORUS064.
Full textThis thesis investigates physiological-based emotion recognition in a digital game context and the feasibility of implementing the model on an embedded system. The following chanllenges are addressed: the relationship between emotional states and physiological responses in the game context, individual variabilities of the pschophysiological responses and issues of implementation on an embedded system. The major contributions of this thesis are : Firstly, we construct a multi-modal Database for Affective Gaming (DAG). This database contains multiple measurements concerning objective modalities: physiological signals (ECG, EDA, EMG, Respiration), screen recording, and player's face recording, as well as subjective assessments on both game event and match level. We presented statistics of the database and run a series of analysis on issues such as emotional moment detection and emotion classification, influencing factors of the overall game experience using various machine learning methods. Secondly, we investigate the individual variability in the collected data by creating an user-specific model and analyzing the optimal feature set for each individual. We proposed a personalized group-based model created the similar user groups by using the clustering techniques based on physiological traits deduced from optimal feature set. We showed that the proposed personalized group-based model performs better than the general model and user-specific model. Thirdly, we implemente the proposed method on an ARM A9 system and showed that the proposed method can meet the requirement of computation time
Ouji, Karima. "Numérisation 3D de visages par une approche de super-résolution spatio-temporelle non-rigide." Phd thesis, Ecole Centrale de Lyon, 2012. http://tel.archives-ouvertes.fr/tel-00923192.
Full textLe, Gall Caroline. "Algorithmes de détection de ruptures et statistiques spatiales : applications au diagnostic de défaillances dans un procédé de fabrication." Toulouse 3, 2002. http://www.theses.fr/2002TOU30176.
Full textThe continuous improvement of the yield of a production line is a significant goal for the competitiveness of the facility. In the context of integrated circuit manufacturing, the introduction of new increasingly complex technologies makes the statistical tools traditionally used insufficient to prevent process failures. Consequently, new statistical techniques have been developed to improve or replace some existing tools and also to form some new ones. Thus, an improvement process is proposed. When a decrease of yield is observed, it first needs to be characterized. The characterization is achieved by a spatial analysis of the silicon wafers on which the integrated circuits are manufactured. . .
Yang, Wenlu. "Personalized physiological-based emotion recognition and implementation on hardware." Electronic Thesis or Diss., Sorbonne université, 2018. https://accesdistant.sorbonne-universite.fr/login?url=https://theses-intra.sorbonne-universite.fr/2018SORUS064.pdf.
Full textThis thesis investigates physiological-based emotion recognition in a digital game context and the feasibility of implementing the model on an embedded system. The following chanllenges are addressed: the relationship between emotional states and physiological responses in the game context, individual variabilities of the pschophysiological responses and issues of implementation on an embedded system. The major contributions of this thesis are : Firstly, we construct a multi-modal Database for Affective Gaming (DAG). This database contains multiple measurements concerning objective modalities: physiological signals (ECG, EDA, EMG, Respiration), screen recording, and player's face recording, as well as subjective assessments on both game event and match level. We presented statistics of the database and run a series of analysis on issues such as emotional moment detection and emotion classification, influencing factors of the overall game experience using various machine learning methods. Secondly, we investigate the individual variability in the collected data by creating an user-specific model and analyzing the optimal feature set for each individual. We proposed a personalized group-based model created the similar user groups by using the clustering techniques based on physiological traits deduced from optimal feature set. We showed that the proposed personalized group-based model performs better than the general model and user-specific model. Thirdly, we implemente the proposed method on an ARM A9 system and showed that the proposed method can meet the requirement of computation time
Kesiman, Made Windu Antara. "Document image analysis of Balinese palm leaf manuscripts." Thesis, La Rochelle, 2018. http://www.theses.fr/2018LAROS013/document.
Full textThe collection of palm leaf manuscripts is an important part of Southeast Asian people’s culture and life. Following the increasing of the digitization projects of heritage documents around the world, the collection of palm leaf manuscripts in Southeast Asia finally attracted the attention of researchers in document image analysis (DIA). The research work conducted for this dissertation focused on the heritage documents of the collection of palm leaf manuscripts from Indonesia, especially the palm leaf manuscripts from Bali. This dissertation took part in exploring DIA researches for palm leaf manuscripts collection. This collection offers new challenges for DIA researches because it uses palm leaf as writing media and also with a language and script that have never been analyzed before. Motivated by the contextual situations and real conditions of the palm leaf manuscript collections in Bali, this research tried to bring added value to digitized palm leaf manuscripts by developing tools to analyze, to transliterate and to index the content of palm leaf manuscripts. These systems aim at making palm leaf manuscripts more accessible, readable and understandable to a wider audience and, to scholars and students all over the world. This research developed a DIA system for document images of palm leaf manuscripts, that includes several image processing tasks, beginning with digitization of the document, ground truth construction, binarization, text line and glyph segmentation, ending with glyph and word recognition, transliteration and document indexing and retrieval. In this research, we created the first corpus and dataset of the Balinese palm leaf manuscripts for the DIA research community. We also developed the glyph recognition system and the automatic transliteration system for the Balinese palm leaf manuscripts. This dissertation proposed a complete scheme of spatially categorized glyph recognition for the transliteration of Balinese palm leaf manuscripts. The proposed scheme consists of six tasks: the text line and glyph segmentation, the glyph ordering process, the detection of the spatial position for glyph category, the global and categorized glyph recognition, the option selection for glyph recognition and the transliteration with phonological rules-based machine. An implementation of knowledge representation and phonological rules for the automatic transliteration of Balinese script on palm leaf manuscript is proposed. The adaptation of a segmentation-free LSTM-based transliteration system with the generated synthetic dataset and the training schemes at two different levels (word level and text line level) is also proposed
Dhouib, Sofiane. "Contributions to unsupervised domain adaptation : Similarity functions, optimal transport and theoretical guarantees." Thesis, Lyon, 2020. http://www.theses.fr/2020LYSEI117.
Full textThe surge in the quantity of data produced nowadays made of Machine Learning, a subfield of Artificial Intelligence, a vital tool used to extract valuable patterns from them and allowed it to be integrated into almost every aspect of our everyday activities. Concretely, a machine learning algorithm learns such patterns after being trained on a dataset called the training set, and its performance is assessed on a different set called the testing set. Domain Adaptation is an active research area of machine learning, in which the training and testing sets are not assumed to stem from the same probability distribution, as opposed to Supervised Learning. In this case, the two distributions generating the training and testing data correspond respectively to the source and target domains. Our contributions focus on three theoretical aspects related to domain adaptation for classification tasks. The first one is learning with similarity functions, which deals with classification algorithms based on comparing an instance to other examples in order to decide its class. The second is large-margin classification, which concerns learning classifiers that maximize the separation between classes. The third is Optimal Transport that formalizes the principle of least effort for transporting probability masses between two distributions. At the beginning of the thesis, we were interested in learning with so-called (epsilon,gamma,tau)-good similarity functions in the domain adaptation framework, since these functions have been introduced in the literature in the classical framework of supervised learning. This is the subject of our first contribution in which we theoretically study the performance of a similarity function on a target distribution, given it is suitable for the source one. Then, we tackle the more general topic of large-margin classification in domain adaptation, with weaker assumptions than those adopted in the first contribution. In this context, we proposed a new theoretical study and a domain adaptation algorithm, which is our second contribution. We derive novel bounds taking the classification margin on the target domain into account, that we convexify by leveraging the appealing Optimal Transport theory, in order to derive a domain adaptation algorithm with an adversarial variation of the classic Kantorovich problem. Finally, after noticing that our adversarial formulation can be generalized to include several other cases of interest, we dedicate our last contribution to adversarial or minimax variations of the optimal transport problem, where we demonstrate the versatility of our approach
Léonardon, Mathieu. "Décodage de codes polaires sur des architectures programmables." Thesis, Bordeaux, 2018. http://www.theses.fr/2018BORD0399/document.
Full textPolar codes are a recently invented class of error-correcting codes that are of interest to both researchers and industry, as evidenced by their selection for the coding of control channels in the next generation of cellular mobile communications (5G). One of the challenges of future mobile networks is the virtualization of digital signal processing, including channel encoding and decoding algorithms. In order to improve network flexibility, these algorithms must be written in software and deployed on programmable architectures.Such a network infrastructure allow dynamic balancing of the computational effort across the network, as well as inter-cell cooperation. These techniques are designed to reduce energy consumption, increase through put and reduce communication latency. The work presented in this manuscript focuses on the software implementation of polar codes decoding algorithms and the design of programmable architectures specialized in their execution.One of the main characteristics of a mobile communication chain is that the state of communication channel changes over time. In order to address issue, adaptive modulationand coding techniques are used in communication standards. These techniques require the decoders to support a wide range of codes : they must be generic. The first contribution of this work is the software implementation of generic decoders for "List" polar decoding algorithms on general purpose processors. In addition to their genericity, the proposed decoders are also flexible. Trade-offs between correction power, throughput and decodinglatency are enabled by fine-tuning the algorithms. In addition, the throughputs of the proposed decoders achieve state-of-the-art performance and, in some cases, exceed it.The second contribution of this work is the proposal of a new high-performance programmable architecture specialized in polar code decoding. It is part of the family of Application Specific Instruction-set Processors (ASIP). The base architecture is a RISC processor. This base architecture is then configured, its instruction set is extended and dedicated hardware units are added. Simulations show that this architecture achieves through puts and latencies close to state-of-the-art software implementations on generalpurpose processors. Energy consumption is reduced by an order of magnitude. The energy required per decoded bit is about 10 nJ on general purpose processors compared to 1nJ on proposed processors when considering the Successive Cancellation (SC) decoding algorithm of a polar code (1024,512).The third contribution of this work is also the design of an ASIP architecture. It differs from the previous one by the use of an alternative design methodology. Instead of being based on a RISC architecture, the proposed processor architecture is part of the classof Transport Triggered Architectures (TTA). It is characterized by a greater modularity that allows to significantly improve the efficiency of the processor. The measured flowrates are then higher than those obtained on general purpose processors. The energy consumption is reduced to about 0.1 nJ per decoded bit for a polar code (1024,512) with the SC decoding algorithm. This corresponds to a reduction of two orders of magnitude compared to the consumption measured on general purpose processors
Ouellet, Sébastien. "Environnement d’adaptation pour un jeu sérieux." Thèse, 2016. http://hdl.handle.net/1866/16179.
Full textWe developed a serious game in order to teach users how to draw Lewis diagrams. We integrated an environment able to record in electroencephalographic signals, facial expressions, and pupil diameters to the serious game. The goal of this work is to determine whether such an environment enabled the serious game to detect in real-time whether or not the user needs help and adapt itself accordingly, and if the experience is more enjoyable for the users if the game tries to adapt itself. Results show that two approaches were promising in order to detect the level of help needed, both training a machine learning models but one using a general data set and the other a personalized (to the user) data set, with their respective performances being 53.4% and 67.5% compared to a chance baseline of 33.3%.
Karabernou, Si Mohamoud. "Conception et réalisation d'un processeur pour une architecture cellulaire massivement parallèle intégrée." Phd thesis, 1992. http://tel.archives-ouvertes.fr/tel-00343216.
Full textSanka, Norbert Bertrand. "Étude comparative et choix optimal du nombre de classes en classification et réseaux de neurones : application en science des données." Thèse, 2021. http://depot-e.uqtr.ca/id/eprint/9662/1/eprint9662.pdf.
Full text