Nous cherchons à équiper un processus d’aide à la décision d’outils permettantde répondre aux exigences de redevabilité. Un décideur fournit de l’information quant à ses préférences au sujet de la façon d’arbitrer entre des points de vue conflictuels. Un analyste, chargé d’éclairer la prise de décision, fait l’hypothèse d’un modèle de raisonnement, et l’ajuste aux informations fournies par le décideur. Nous faisons l’hypothèse d’un processus d’élicitation robuste, dont les recommandations sont déduites des éléments dialectiques. Nous nous sommes donc intéressés à la résolution d’un problème inverse concernant le modèle, ainsi qu’à la production d’explications, si possible correctes, complètes, facile à calculer et à comprendre. Nous avons considéré deux formes de représentation du raisonnement: l’une ayant trait à la comparaison de paires d’alternatives fondée sur un modèle de valeur additive, l’autre ayant trait au tri des alternatives dans des catégories ordonnées fondé sur un raisonnement non-compensatoire. / We consider providing a decision aiding process with tools aiming at complying to the demands of accountability. Decision makers, seeking support, provide preference information in the form of reference cases, that illustrates their views on the way of taking into account conflicting points of view. The analyst, who provides the support, assumes a generic representation of the reasoning with preferences, and fits the aggregation procedure to the preference information. We assume a robust elicitation process, where the recommendations stemming from the fitted procedure can be deduced from dialectical elements. Therefore, we are interested in solving an inverse problem concerning the model, and in deriving explanations, if possible sound, complete, easy to compute and to understand. We address two distinct forms of reasoning: one aimed at comparing pairs of alternatives with an additive value model, the other aimed at sorting alternatives into ordered categories with a noncompensatory model.

Description logics (DL) knowledge bases are often build by users with expertise in the application domain, but little expertise in logic. To support this kind of users when building their knowledge bases a number of extension methods have been proposed to provide the user with concept descriptions as a starting point for new concept definitions. The inference service central to several of these approaches is the computation of (least) common subsumers of concept descriptions. In case disjunction of concepts can be expressed in the DL under consideration, the least common subsumer (lcs) is just the disjunction of the input concepts. Such a trivial lcs is of little use as a starting point for a new concept definition to be edited by the user. To address this problem we propose two approaches to obtain "meaningful" common subsumers in the presence of disjunction tailored to two different methods to extend DL knowledge bases. More precisely, we devise computation methods for the approximation-based approach and the customization of DL knowledge bases, extend these methods to DLs with number restrictions and discuss their efficient implementation.

In the current age of new textual information emerging constantly, there is a challenge related to processing and structuring it in some ways. Moreover, the information is often expressed in many different languages, but the discourse tends to be dominated by English, which may lead to overseeing important, specific knowledge in less well-resourced languages. Knowledge graphs have been proposed as a way of structuring unstructured data, making it machine-readable and available for further processing. Researchers have also emphasized the potential bilateral benefits of combining knowledge in low- and well-resourced languages.  In this thesis, I combine the two goals of structuring textual data with the help of knowledge graphs and including multilingual information in an effort to achieve a more accurate knowledge representation. The purpose of the project is to investigate whether the information about three Swedish companies known worldwide - H&M, Spotify, and Ikea - in Swedish and English data sources is the same and how combining the two sources can be beneficial. Following a natural language processing (NLP) pipeline consisting of such tasks as coreference resolution, entity linking, and relation extraction, a knowledge graph is constructed from Swedish and English news articles about the companies. Refinement techniques are applied to improve the graph. The constructed knowledge graph is analyzed with respect to the overlap of extracted entities and the complementarity of information. Different variants of the graph are further evaluated by human raters. A number of queries illustrate the capabilities of the constructed knowledge graph. The evaluation of the graph shows that the topics covered in the two information sources differ substantially. Only a small number of entities occur in both languages. Combining the two sources can, therefore, contribute to a richer and more connected knowledge graph. The adopted refinement techniques increase the connectedness of the graph. Human evaluators consequently chose the Swedish side of the data as more relevant for the considered questions, which points out the importance of not limiting the research to more easily available and processed English data.

<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>

<div>By Machine Learning in the Open World, we are trying to build models that can be used in a more realistic setting where there could always be something "unknown" happening. Beyond the traditional machine learning tasks such as classification and segmentation where all classes are predefined, we are dealing with the challenges from newly emerged classes, irrelevant classes, outliers, and class imbalance.</div><div>At the beginning, we focus on the Non-Exhaustive Learning (NEL) problem from a statistical aspect. By NEL, we assume that our training classes are non-exhaustive, where the testing data could contain unknown classes. And we aim to build models that could simultaneously perform classification and class discovery. We proposed a non-parametric Bayesian model that learns some hyper-parameters from both training and discovered classes (which is empty at the beginning), then infer the label partitioning under the guidance of the learned hyper-parameters, and repeat the above procedure until convergence.</div><div>After obtaining good results on applications with plain and low dimensional data such flow-cytometry and some benchmark datasets, we move forward to Non-Exhaustive Feature Learning (NEFL). For NEFL, we extend our work with deep learning techniques to learn representations on datasets with complex structural and spatial correlations. We proposed a metric learning approach to learn a feature space with good discrimination on both training classes and generalize well on unknown classes. Then we developed some variants of this metric learning algorithm to deal with outliers and irrelevant classes. We applied our final model to applications such as open world image classification, image segmentation, and SRS hyperspectral image segmentation and obtained promising results.</div><div>Finally, we did some explorations with Out of Distribution detection (OOD) to detect irrelevant sample and outliers to complete the story.</div>

<p>Real-world datasets are rarely clean. This causes the discrepancy between the claimed performance of machine learning algorithms on paper, and their actual performance on real-world problems. When dealing with missing or hidden information in a dataset, researchers have been using heuristic imputation methods since the first day of machine learning. However, it is known that many imputation methods do not have theoretical guarantees in various machine learning tasks, including clustering, community detection, sparsity recovery, to name a few. On the other hand, theoretical machine learning papers often follow simplistic assumptions, which are rarely fulfilled in real-world datasets. My research focuses on developing statistically and computationally efficient learning algorithms with provable guarantees under novel incomplete and latent assumptions. We consider problems with arguably more realistic incomplete and latent assumptions.We provide analysis to community detection in various network models, inference with latent variables in an arbitrary planted model, federated myopic community detection, and high-order tensor models. We analyze the interaction between the missing or latent structures and the inference / recoverability conditions, and proposed algorithms to solve the problems efficiently. <br> <br> Our main contributions in this thesis are as follows.<br> </p> <ol> <li>We analyze the information-theoretic limits for the recovery of node labels in several network models. We analyze the information-theoretic limits for community detection. We carefully construct restricted ensembles for a subclass of network models, and provide a series of novel results. </li> <li>We analyze the necessary and sufficient conditions for exact inference of a latent model. We show that exact inference can be achieved using a semidefinite programming approach without knowing either the latent variables or their domain. Our analysis predicts the experimental correctness of SDP with high accuracy, showing the suitability of our focus on the Karush-Kuhn-Tucker conditions and the spectrum of a properly defined matrix.</li> <li>We study the problem of recovering the community structure of a network under federated myopic learning. Under this paradigm, we have several clients, each of them having a myopic view, i.e., observing a small subgraph of the network. Each client sends a censored evidence graph to a central server. We provide an efficient algorithm, which computes a consensus signed weighted graph from clients evidence, and recovers the underlying network structure in the central server. We analyze the topological structure conditions of the network, as well as the signal and noise levels of the clients that allow for recovery of the network structure. Our analysis shows that exact recovery is possible and can be achieved in polynomial time.</li> <li>We study the problem of exact partitioning of high-order models. We consider two different high-order assumptions, and show that exact partitioning of high-order planted models is achievable through solving a convex optimization problem with a novel Carathéodory symmetric tensor cone in one case, and with a tensor nuclear norm constraint in the other.</li> <li>We study the problem of inference in high-order structured prediction tasks. We apply a generative model approach to study the problem of high-order inference, and provide a two-stage convex optimization algorithm for exact label recovery. We also connect the performance of our algorithm and the hyperedge expansion property using a novel hypergraph Cheeger-type inequality.</li> <li>We study the problem of partial recovery through semidefinite programming. We are interested in the scenarios in which the SDP returns a solution that is partially correct without any rounding. We analyze the optimality condition of partial recovery and provide statistical and topological guarantees. </li> </ol>

Question answering (QA) stands squarely along the path from document retrieval to text understanding. As an area of research interest, it serves as a proving ground where strategies for document processing, knowledge representation, question analysis, and answer extraction may be evaluated in real world information extraction contexts. The task is to go beyond the representation of text documents as "bags of words" or data blobs that can be scanned for keyword combinations and word collocations in the manner of internet search engines. Instead, the goal is to recognize and extract the semantic content of the text, and to organize it in a manner that supports reasoning about the concepts represented. The issue presented is how to obtain and query such a structure without either a predefined set of concepts or a predefined set of relationships among concepts. This research investigates a means for acquiring from text documents both the underlying concepts and their interrelationships. Specifically, a syntax-based formalism for representing atomic propositions that are extracted from text documents is presented, together with a method for constructing a network of concept nodes for indexing such logical forms based on the discourse entities they contain. It is shown that meaningful questions can be decomposed into Boolean combinations of question patterns using the same formalism, with free variables representing the desired answers. It is further shown that this formalism can be used for robust question answering using the concept network and WordNet synonym, hypernym, hyponym, and antonym relationships. This formalism was implemented in the Semantic Extractor (SEMEX) research tool and was tested against the factoid questions from the 2005 Text Retrieval Conference (TREC), which operated upon the AQUAINT corpus of newswire documents. After adjusting for the limitations of the tool and the document set, correct answers were found for approximately fifty percent of the questions analyzed, which compares favorably with other question answering systems.

An important goal for machine learning is to transfer knowledge between tasks. For example, learning to play RoboCup Keepaway should contribute to learning the full game of RoboCup soccer. Often approaches to task transfer focus on transforming the original representation to fit the new task. Such representational transformations are necessary because the target task often requires new state information that was not included in the original representation. In RoboCup Keepaway, changing from the 3 vs. 2 variant of the task to 4 vs. 3 adds state information for each of the new players. In contrast, this dissertation explores the idea that transfer is most effective if the representation is designed to be the same even across different tasks. To this end, (1) the bird’s eye view (BEV) representation is introduced, which can represent different tasks on the same two-dimensional map. Because the BEV represents state information associated with positions instead of objects, it can be scaled to more objects without manipulation. In this way, both the 3 vs. 2 and 4 vs. 3 Keepaway tasks can be represented on the same BEV, which is (2) demonstrated in this dissertation. Yet a challenge for such representation is that a raw two-dimensional map is highdimensional and unstructured. This dissertation demonstrates how this problem is addressed naturally by the Hypercube-based NeuroEvolution of Augmenting Topologies (HyperNEAT) approach. HyperNEAT evolves an indirect encoding, which compresses the representation by exploiting its geometry. The dissertation then explores further exploiting the power of such encoding, beginning by (3) enhancing the configuration of the BEV with a focus on iii modularity. The need for further nonlinearity is then (4) investigated through the addition of hidden nodes. Furthermore, (5) the size of the BEV can be manipulated because it is indirectly encoded. Thus the resolution of the BEV, which is dictated by its size, is increased in precision and culminates in a HyperNEAT extension that is expressed at effectively infinite resolution. Additionally, scaling to higher resolutions through gradually increasing the size of the BEV is explored. Finally, (6) the ambitious problem of scaling from the Keepaway task to the Half-field Offense task is investigated with the BEV. Overall, this dissertation demonstrates that advanced representations in conjunction with indirect encoding can contribute to scaling learning techniques to more challenging tasks, such as the Half-field Offense RoboCup soccer domain.

Skill acquisition and task specific planning are essential components of any robot system, yet they have long been studied in isolation. This, I contend, is due to the lack of a common representational framework. I present a holistic approach to planning robot behavior, using previously acquired skills to represent control knowledge (and objects) directly, and to use this background knowledge to build plans in the space of control actions. Actions in this framework are closed-loop controllers constructed from combinations of sensors, effectors, and potential functions. I will show how robots can use reinforcement learning techniques to acquire sensorimotor programs. The agent then builds a functional model of its interactions with the world as distributions over the acquired skills. In addition, I present two planning algorithms that can reason about a task using the functional models. These algorithms are then applied to a variety of tasks such as object recognition and object manipulation to achieve its objective on two different robot platforms.

No description available.