No description available.

No description available.

京都大学 / 新制・課程博士 / 博士(人間健康科学) / 甲第24539号 / 人健博第110号 / 新制||人健||8(附属図書館) / 京都大学大学院医学研究科人間健康科学系専攻 / (主査)教授 木下 彩栄, 教授 中尾 恵, 教授 中山 健夫 / 学位規則第4条第1項該当 / Doctor of Human Health Sciences / Kyoto University / DFAM

<div>Building machine learning algorithms to recognize objects in real-world tasks is a very challenging problem. With increasing number of classes, it becomes very costly and impractical to collect samples for all classes to obtain an exhaustive data to train the model. This limited labeled data bottleneck prevails itself more profoundly over fine grained object classes where some of these classes may lack any labeled representatives in the training data. A robust algorithm in this realistic scenario will be required to classify samples from well-represented classes as well as to handle samples from unknown origin. In this thesis, we break down this difficult task into more manageable sub-problems and methodically explore novel solutions to address each component in a sequential order.</div><div><br></div><div>We begin with zero-shot learning (ZSL) scenario where classes that are lacking any labeled images in the training data, i.e., unseen classes, are assumed to have some semantic descriptions associated with them. The ZSL paradigm is motivated by analogy to humans’ learning process. We human beings can recognize new categories by just knowing some semantic descriptions of them without even seeing any instances from these categories. We</div><div>develop a novel hierarchical Bayesian classifier for ZSL task. The two-layer architecture of the model is specifically designed to exploit the implicit hierarchy present among classes, in particular evident in fine-grained datasets. In the proposed method, there are latent classes that define the class hierarchy in the image space and semantic information is used to build the Bayesian hierarchy around these meta-classes. Our Bayesian model imposes local priors on semantically similar classes that share the same meta-class to realize knowledge transfer. We finally derive posterior predictive distributions to reconcile information about local and global priors and then blend them with data likelihood for the final likelihood calculation. With its closed form solution, our two-layer hierarchical classifier proves to be fast in training and flexible to model both fine and coarse-grained datasets. In particular, for challenging fine-grained datasets the proposed model can leverage the large number of seen classes to its advantage for a better local prior estimation without sacrificing on seen class accuracy.</div><div>Side information plays a critical role in ZSL and ZSL models hold on a strong assumption that the side information is strongly correlated with image features. Our model uses side information only to build hierarchy, thus, no explicit correlation between image features is assumed. This in turn leads the Bayesian model to be very resilient to various side</div><div>information sources as long as they are discriminative enough to define class hierarchy.</div><div><br></div><div>When dealing with thousands of classes, it becomes very difficult to obtain semantic descriptions for fine grained classes. For example, in species classification where classes display very similar morphological traits, it is impractical if not impossible to derive characteristic</div><div>visual attributes that can distinguish thousands of classes. Moreover, it would be unrealistic to assume that an exhaustive list of visual attributes characterizing all object classes, both seen and unseen, can be determined based only on seen classes. We propose DNA as a side</div><div>information to overcome this obstacle in order to do fine grained zero-shot species classification. We demonstrate that 658 base pair long DNA barcodes can be sufficient to serve as a robust source of side information for newly compiled insect dataset with more than thousand</div><div>classes. The experiments is further validated on well-known CUB dataset on which DNA attributes proves to be as competitive as word vectors. Our proposed Bayesian classifier delivers state of the art results on both datasets while using DNA as side information.</div><div><br></div><div>Traditional ZSL framework, however, is not quite suitable for scalable species identification and discovery. For example, insects are one of the largest groups of animal kingdom</div><div>with estimated 5.5 million species yet only 20% of them is described. We extend the traditional ZSL into a more practical framework where no explicit side information is available for unseen classes. We transform our Bayesian model to utilize taxonomical hierarchy of species</div><div>to perform insect identification at scale. Our approach is the first to combine two different data modalities, namely image and DNA information, to perform insect identification with</div><div>more than thousand classes. Our algorithm not only classifies known species with impressive 97% accuracy but also identifies unknown species and classify them to their true genus with 81% accuracy.</div><div><br></div><div>Our approach has the ability to address some major societal issues in climate change such as changing insect distributions and measuring biodiversity across the world. We believe this work can pave the way for more precise and more importantly the scalable monitoring of</div><div>biodiversity and can become instrumental in offering objective measures of the impacts of recent changes our planet has been going through.</div>

The video-based computational analyses of human face and gesture signals encompass a myriad of challenging research problems involving computer vision, machine learning and human computer interaction. In this thesis, we focus on the following challenges: a) the classification of hand and body gestures along with the temporal localization of their occurrence in a continuous stream, b) the recognition of facial expressivity levels in people with Parkinson's Disease using multimodal feature representations, c) the prediction of student learning outcomes in intelligent tutoring systems using affect signals, and d) the personalization of machine learning models, which can adapt to subject and group-specific nuances in facial and gestural behavior. Specifically, we first conduct a quantitative comparison of two approaches to the problem of segmenting and classifying gestures on two benchmark gesture datasets: a method that simultaneously segments and classifies gestures versus a cascaded method that performs the tasks sequentially. Second, we introduce a framework that computationally predicts an accurate score for facial expressivity and validate it on a dataset of interview videos of people with Parkinson's disease. Third, based on a unique dataset of videos of students interacting with MathSpring, an intelligent tutoring system, collected by our collaborative research team, we build models to predict learning outcomes from their facial affect signals. Finally, we propose a novel solution to a relatively unexplored area in automatic face and gesture analysis research: personalization of models to individuals and groups. We develop hierarchical Bayesian neural networks to overcome the challenges posed by group or subject-specific variations in face and gesture signals. We successfully validate our formulation on the problems of personalized subject-specific gesture classification, context-specific facial expressivity recognition and student-specific learning outcome prediction. We demonstrate the flexibility of our hierarchical framework by validating the utility of both fully connected and recurrent neural architectures.

No description available.

No description available.

No description available.

L'estimation des abondances de population est essentielle pour comprendre les dynamiques de population, les interactions entre espèces et estimer les risques de transmission d'agents pathogènes dans les populations. Plusieurs méthodes d'échantillonnages, basées sur des hypothèses spécifiques permettent d'estimer ces abondances : les méthodes par comptages uniques, par « distance sampling », par échantillonnages successifs ou par capture marquage recapture. Nous nous sommes intéressés à l'abondance des tiques Ixodes ricinus, vecteurs de nombreux agents pathogènes. Cette abondance est classiquement estimée par le nombre de tiques capturées lors d'échantillonnages uniques réalisés sur différentes unités d'observation. Cependant, de nombreuses études remettent en cause cette hypothèse forte et suggèrent que le taux d'échantillonnage est variable selon les conditions d'échantillonnage (type de végétation,…) mais ne prennent pas en compte ce taux d'échantillonnage pour autant. A partir d'une méthode d'échantillonnage par « removal sampling » (RS), (i) nous avons montré que les conditions environnementales influençaient le taux d'échantillonnage et l'indicateur d'abondance usuel i.e. le nombre de tiques capturées lors d'un seul échantillonnage (ii) nous avons proposé une méthode pour détecter l'indicateur d'abondance, basés sur le nombre cumulé de capture, le moins soumis aux variations du taux ; (iii) par une approche Bayésienne hiérarchique, nous avons estimé simultanément l'abondance de tiques des unités d'observation et la valeur du taux d'échantillonnage en fonction du type de végétation et de l'heure d'échantillonnage. Nous avons montré que le taux d'échantillonnage sur des arbustes (entre 33,9 % et 47,4%) était significativement inférieur au taux d'échantillonnage sur des feuilles mortes (entre 53,6 % et 66,7%). De plus, nous avons montré que le modèle RS tend vers un modèle de Poisson iid lorsque la taille de la population N0 tend vers l'infini ce qui pose des problèmes d'indétermination pour estimer les paramètres N0 et τ, le taux d'échantillonnage. Nous avons également montré que (i) les estimateurs Bayésiens divergent lorsque les lois a priori sont des lois vagues ; (ii) les lois a priori β(a, b) avec a > 2 sur τ conduisaient à des estimateurs Bayésien convergents. Enfin, nous avons proposé des recommandations quant au choix des lois a priori pour τ afin d'obtenir de bonnes estimations pour N0 ou pour τ. Nous discutons de la pertinence des méthodes RS pour les tiques et des perspectives envisageables pour (i) estimer le risque acarologique représenté par la population de tiques potentiellement actives sur une unité d'observation, (ii) estimer un risque à l'échelle d'une parcelle, à savoir comment répartir l'effort d'échantillonnage entre le nombre d'unités d'observation et le nombre d'échantillonnages successifs par unités d'observation. / The estimation of animal abundance is essential to understand population dynamics, species interactions and disease patterns in populations and to estimate the risk of pathogens transmission. Several sampling methods such as single counts, distance sampling, removal sampling or capture mark recapture could be used to estimate abundance. In this study, we are investigated the abundance of Ixodes ricinus ticks, which are involved in the transmission of many pathogens. Tick abundance is commonly estimated by the number of nymphs captured during a single observation (a cloth dragged on a given surface). In this case, analyses of abundance patterns assumes that the probability of detecting a tick, hence the sampling rate, remains constant across the observations. In practice, however, this assumption is often not satisfied as the sampling rate may fluctuate between observation plots. The variation of sampling rate is never taken into account in estimations of tick abundance. Using a removal sampling design (RS), (i) we showed that the sampling rate and the usual abundance indicator (based on a single drag observation per spot) were both influenced by environmental conditions ; (ii) we proposed a method to determine the abundance indicator the least influenced by sampling rate variations ; (iii) using a hierarchical Bayesian model, we estimated simultaneously the abundance and the sampling rate according the type of vegetation, and the time of sampling. The sampling rate varied between 33,9 % and 47,4 % for shrubs and 53,6 % and 66,7 % for dead leaves. In addition, we show that the RS model tends to Poisson iid model when the population size N0 tends to infinite. This result conduct to infinite estimations for N0. We show that (i) Bayesian estimators were divergent for vague prior ; (ii) β(a, b) prior for a > 2 on τ conduct to convergent estimators. Then, we proposed recommendations for prior choice for τ parameter to give good estimations of N0 or τ. We discuss the relevance of RS for ticks and the possible perspectives to (i) estimate the acarologic risk associated to all potential active ticks for given spot, (ii) estimate the risk at the larger scale, i.e. how to distribute the sampling effort between number of spot and number of consecutive sampling by spot.

Un essai clinique est une recherche biomédicale pratiquée sur l'Homme dont l'objectif est la consolidation et le perfectionnement des connaissances biologiques ou médicales. Le nombre de sujets nécessaire (NSN) est le nombre minimal de patients à inclure dans l'essai afin d'assurer au test statistique une puissance donnée pour observer un effet donné. Pour ce faire plusieurs centres investigateurs sont sollicités. La période entre l'ouverture du premier centre investigateur et le recrutement du dernier patient est appelée période de recrutement que l'on souhaite modéliser. Les premières modélisations remontent à presque 50 ans avec les travaux de Lee, Williford et al. et Morgan avec l'idée déjà d'une modélisation de la dynamique de recrutement par des processus de Poisson. Un problème émerge lors de recrutement multicentriques du fait du manque de caractérisation de l'ensemble des sources de variabilité agissant sur les différentes dynamiques de recrutement. Le modèle dit Poisson-gamma basé sur un processus de Poisson dont les intensités par centre sont considérées comme un échantillon de loi gamma permet l'étude de variabilité. Ce modèle est au coeur de notre projet. Différents objectifs ont motivés la réalisation de cette thèse. Le premier questionnement porte sur la validité de ces modèles. Elle est établie de façon asymptotique et une étude par simulation permet de donner des informations précises sur la validité du modèle. Par la suite l'analyse de bases de données réelles a permis de constater que lors de certaines phases de recrutement, des pauses dans le recrutement sont observables. Une question se pose alors naturellement : comment et faut-il prendre en compte ces informations dans le modèle de dynamique de recrutement ? Il résulte d'études par simulation que la prise en compte de ces données n'améliore pas les performances prédictives du modèle lorsque les sources d'interruptions sont aléatoires mais dont la loi est inchangée au cours du temps. Une autre problématique observable sur les données et inhérente au problème de recrutement de patients est celle des dites sorties d'étude. Une technique Bayésienne empirique analogue à celle du processus de recrutement peut être introduite pour modéliser les sorties d'étude. Ces deux modélisations se couplent très bien et permettent d'estimer la durée de recrutement ainsi que la probabilité de sorties d'étude en se basant sur les données de recrutement d'une étude intermédiaire, donnant des prédictions concernant le processus de randomisation. La dynamique de recrutement possède de multiples facteurs autre que le temps de recrutement. Ces aspects fondamentaux couplés au modèle Poisson-gamma fournissent des indicateurs pertinents pour le suivi des essais. Ainsi est-il possible d'ajuster le nombre de centres au cours de l'essai en fonction d'objectifs prédéfinis, de modéliser et prévoir la chaîne d'approvisionnement nécessaire lors de l'essai et de prévoir l'effet de la randomisation des patients par région sur la puissance du test de l'essai. Il permet également d'avoir un suivi des patients après randomisation permettant ainsi de prévoir un ajustement du nombre de patients en cas de pertes significative d'effectif, ou d'abandonner un essai si les résultats préliminaires sont trop faibles par rapport aux risques connus et observés. La problématique de la dynamique de recrutement peut être couplée avec la dynamique de l'étude en elle-même quand celle-ci est longitudinale. L'indépendance des deux processus permet une estimation facile des différents paramètres. Le résultat est un modèle global du parcours du patient dans l'essai. Deux exemples clés de telles situations sont les données de survie - la modélisation permet alors d'estimer la durée d'un essai quand le critère d'arrêt est le nombre d'événements observés et les modèles de Markov - la modélisation permet alors d'estimer le nombre de patients dans un certain état au bout d'un certain temps. / A clinical trial is a biomedical research which aims to consolidate and improve the biological and medical knowledges. The number of patients required il the minimal number of patients to include in the trial in order to insure a given statistical power of a predefined test. The constitution of this patients' database is one of the fundamental issues of a clinical trial. To do so several investigation centres are opened. The duration between the first opening of a centre and the last recruitment of the needed number of patients is called the recruitemtn duration that we aim to model. The fisrt model goes back 50 years ago with the work of Lee, Williford et al. and Morgan with the idea to model the recruitment dynamic using Poisson processes. One problem emerge, that is the lack of caracterisation of the variabliity of recruitment between centers that is mixed with the mean of the recruitment rates. The most effective model is called the Poisson-gamma model which is based on Poisson processes with random rates (Cox process) with gamma distribution. This model is at the very heart of this project. Different objectives have motivated the realisation of this thesis. First of all the validity of the Poisson-gamma model is established asymptotically. A simulation study that we made permits to give precise informations on the model validity in specific cases (function of the number of centers, the recruitement duration and the mean rates). By studying database, one can observe that there can be breaks during the recruitment dynamic. A question that arise is : How and must we take into account this phenomenon for the prediction of the recruitment duration. The study made tends to show that it is not necessary to take them into account when they are random but their law is stable in time. It also veered around to measure the impact of these breaks on the estimations of the model, that do not impact its validity under some stability hypothesis. An other issue inherent to a patient recruitment dynamic is the phenomenon of screening failure. An empirical Bayesian technique analogue to the one of the recruitment process is used to model the screening failure issue. This hierarchical Bayesian model permit to estimate the duartion of recruitment with screening failure consideration as weel as the probability to drop out from the study using the data at some interim time of analysis, giving predictions on the randomisation dynamic. The recruitment dynamic can be studied in many different ways than just the duration of recruitment. These fundamental aspects coupled with the Poisson-gamma model give relevant indicators for the study follow-up. Multiples applications in this sense are computed. It is therefore possible to adjust the number of centers according to predefined objectives, to model the drug's supply chain per region or center and to predict the effect of the randomisation on the power of the test's study. It also allows to model the folow-up period of the patients by means of transversal or longitudinal methods, that can serve to adjust the number of patients if too many quit during the foloww-up period, or to stop the study if dangerous side effects or no effects are observed on interim data. The problematic of the recruitment dynamic can also be coupled with the dynamic of the study itself when it is longitudinal. The independance between these two processes allows easy estimations of the different parameters. The result is a global model of the patient pathway in the trail. Two key examples of such situations are survival data - the model permit to estimate the duration of the trail when the stopping criterion is the number of events observed, and the Markov model - the model permit to estimate the number of patients in a certain state for a given duartion of analysis.