Learning representations for Information Retrieval

La recherche d'informations s'intéresse, entre autres, à répondre à des questions comme: est-ce qu'un document est pertinent à une requête ? Est-ce que deux requêtes ou deux documents sont similaires ? Comment la similarité entre deux requêtes ou documents peut être utilisée pour améliorer l'estimation de la pertinence ? Pour donner réponse à ces questions, il est nécessaire d'associer chaque document et requête à des représentations interprétables par ordinateur. Une fois ces représentations estimées, la similarité peut correspondre, par exemple, à une distance ou une divergence qui opère dans l'espace de représentation. On admet généralement que la qualité d'une représentation a un impact direct sur l'erreur d'estimation par rapport à la vraie pertinence, jugée par un humain. Estimer de bonnes représentations des documents et des requêtes a longtemps été un problème central de la recherche d'informations. Le but de cette thèse est de proposer des nouvelles méthodes pour estimer les représentations des documents et des requêtes, la relation de pertinence entre eux et ainsi modestement avancer l'état de l'art du domaine. Nous présentons quatre articles publiés dans des conférences internationales et un article publié dans un forum d'évaluation. Les deux premiers articles concernent des méthodes qui créent l'espace de représentation selon une connaissance à priori sur les caractéristiques qui sont importantes pour la tâche à accomplir. Ceux-ci nous amènent à présenter un nouveau modèle de recherche d'informations qui diffère des modèles existants sur le plan théorique et de l'efficacité expérimentale. Les deux derniers articles marquent un changement fondamental dans l'approche de construction des représentations. Ils bénéficient notamment de l'intérêt de recherche dont les techniques d'apprentissage profond par réseaux de neurones, ou deep learning, ont fait récemment l'objet. Ces modèles d'apprentissage élicitent automatiquement les caractéristiques importantes pour la tâche demandée à partir d'une quantité importante de données. Nous nous intéressons à la modélisation des relations sémantiques entre documents et requêtes ainsi qu'entre deux ou plusieurs requêtes. Ces derniers articles marquent les premières applications de l'apprentissage de représentations par réseaux de neurones à la recherche d'informations. Les modèles proposés ont aussi produit une performance améliorée sur des collections de test standard. Nos travaux nous mènent à la conclusion générale suivante: la performance en recherche d'informations pourrait drastiquement être améliorée en se basant sur les approches d'apprentissage de représentations.

Digital Classics Outside the Echo-Chamber : Teaching, Knowledge Exchange and Public Engagement /

The international perspectives on these issues are especially valuable in an increasingly connected, but still institutionally and administratively diverse world. The research addressed in several chapters in this volume includes issues around technical standards bodies like EpiDoc and the TEI, engaging with ways these standards are implemented, documented, taught, used in the process of transcribing and annotating texts, and used to generate publications and as the basis for advanced textual or corpus research. Other chapters focus on various aspects of philological research and content creation, including collaborative or community driven efforts, and the issues surrounding editorial oversight, curation, maintenance and sustainability of these resources. Research into the ancient languages and linguistics, in particular Greek, and the language teaching that is a staple of our discipline, are also discussed in several chapters, in particular for ways in which advanced research methods can lead into language technologies and vice versa and ways in which the skills around teaching can be used for public engagement, and vice versa. A common thread through much of the volume is the importance of open access publication or open source development and distribution of texts, materials, tools and standards, both because of the public good provided by such models (circulating materials often already paid for out of the public purse), and the ability to reach non-standard audiences, those who cannot access rich university libraries or afford expensive print volumes. Linked Open Data is another technology that results in wide and free distribution of structured information both within and outside academic circles, and several chapters present academic work that includes ontologies and RDF, either as a direct research output or as essential part of the communication and knowledge representation. Several chapters focus not on the literary and philological side of classics, but on the study of cultural heritage, archaeology, and the material supports on which original textual and artistic material are engraved or otherwise inscribed, addressing both the capture and analysis of artefacts in both 2D and 3D, the representation of data through archaeological standards, and the importance of sharing information and expertise between the several domains both within and without academia that study, record and conserve ancient objects. Almost without exception, the authors reflect on the issues of interdisciplinarity and collaboration, the relationship between their research practice and teaching and/or communication with a wider public, and the importance of the role of the academic researcher in contemporary society and in the context of cutting edge technologies. How research is communicated in a world of instant- access blogging and 140-character micromessaging, and how our expectations of the media affect not only how we publish but how we conduct our research, are questions about which all scholars need to be aware and self-critical.

Approaches and Strategy for Cancer Research and Surveillance Data: Integration, Information Pipeline, Data Models, and Informatics Opportunities

Thesis (Ph.D.)--University of Washington, 2016-04

New Techniques in Deep Representation Learning

Thesis (Ph.D.)--University of Washington, 2016-06

Graph-based Semi-Supervised Learning in Acoustic Modeling for Automatic Speech Recognition

Thesis (Ph.D.)--University of Washington, 2016-06

Simulating Nursing: Rhetoric, Materiality, and Disciplinary Learning

Thesis (Ph.D.)--University of Washington, 2016-06

Dynamic Models under Uncertainty for Large-Scale Enterprise Systems

Thesis (Ph.D.)--University of Washington, 2016-06

Integrating science and management through collaborative learning and better information management

Achieving more sustainable land and water use depends on high-quality information and its improved use. In other words, better linkages are needed between science and management. Since many stakeholders with different relationships to the natural resources are inevitably involved, we suggest that collaborative learning environments and improved information management are prerequisites for integrating science and management. Case studies that deal with resource management issues are presented that illustrate the creation of collaborative learning environments through systems analyses with communities, and an integration of scientific and experiential knowledge of components of the system. This new knowledge needs to be captured and made accessible through innovative information management systems designed collaboratively with users, in forms which fit the users' 'mental models' of how their systems work. A model for linking science and resource management more effectively is suggested. This model entails systems thinking in a collaborative learning environment, and processes to help convergence of views and value systems, and make scientists and different kinds of managers aware of their interdependence. Adaptive management provides a mechanism for applying and refining scientists' and managers' knowledge. Copyright (C) 2003 John Wiley Sons, Ltd.

Learning of foraging skills by fish

This chapter outlines the relationships between a number of key factors that influence learning and memory, and illustrates them by reference to studies on the foraging behaviour of fish. Learning can lead to significant improvements in foraging performance in only a few exposures, and at least some fish species are capable of adjusting their foraging strategy as patterns of patch profitability change. There is also evidence that the memory window for prey varies between fish species, and that this may be a function of environmental predictability. Convergence between behavioural ecology and comparative psychology offers promise in terms of developing more mechanistically realistic foraging models and explaining apparently 'suboptimal' patterns of behaviour. Foraging decisions involve the interplay between several distinct systems of learning and memory, including those that relate to habitat, food patches, prey types, conspecifics and predators. Fish biologists, therefore, face an interesting challenge in developing integrated accounts of fish foraging that explain how cognitive sophistication can help individual animals to deal with the complexity of the ecological context.

A conceptual model of personalized virtual learning environments

The Virtual Learning Environment (VLE) is one of the fastest growing areas in educational technology research and development. In order to achieve learning effectiveness, ideal VLEs should be able to identify learning needs and customize solutions, with or without an instructor to supplement instruction. They are called Personalized VLEs (PVLEs). In order to achieve PVLEs success, comprehensive conceptual models corresponding to PVLEs are essential. Such conceptual modeling development is important because it facilitates early detection and correction of system development errors. Therefore, in order to capture the PVLEs knowledge explicitly, this paper focuses on the development of conceptual models for PVLEs, including models of knowledge primitives in terms of learner, curriculum, and situational models, models of VLEs in general pedagogical bases, and particularly, the definition of the ontology of PVLEs on the constructivist pedagogical principle. Based on those comprehensive conceptual models, a prototyped multiagent-based PVLE has been implemented. A field experiment was conducted to investigate the learning achievements by comparing personalized and non-personalized systems. The result indicates that the PVLE we developed under our comprehensive ontology successfully provides significant learning achievements. These comprehensive models also provide a solid knowledge representation framework for PVLEs development practice, guiding the analysis, design, and development of PVLEs. (c) 2005 Elsevier Ltd. All rights reserved.

Applying machine learning to catalogue matching in astrophysics

We present the results of applying automated machine learning techniques to the problem of matching different object catalogues in astrophysics. In this study, we take two partially matched catalogues where one of the two catalogues has a large positional uncertainty. The two catalogues we used here were taken from the H I Parkes All Sky Survey (HIPASS) and SuperCOSMOS optical survey. Previous work had matched 44 per cent (1887 objects) of HIPASS to the SuperCOSMOS catalogue. A supervised learning algorithm was then applied to construct a model of the matched portion of our catalogue. Validation of the model shows that we achieved a good classification performance (99.12 per cent correct). Applying this model to the unmatched portion of the catalogue found 1209 new matches. This increases the catalogue size from 1887 matched objects to 3096. The combination of these procedures yields a catalogue that is 72 per cent matched.

An examination of learning processes during critical incident training: Implications for the development of adaptable trainees

Three experiments are reported that examined the process by which trainees learn decision-making skills during a critical incident training program. Formal theories of category learning were used to identify two processes that may be responsible for the acquisition of decision-making skills: rule learning and exemplar learning. Experiments I and 2 used the process dissociation procedure (L. L. Jacoby, 1998) to evaluate the contribution of these processes to performance. The results suggest that trainees used a mixture of rule and exemplar learning. Furthermore, these learning processes were influenced by different aspects of training structure and design. The goal of Experiment 3 was to develop training techniques that enable trainees to use a rule adaptively. Trainees were tested on cases that represented exceptions to the rule. Unexpectedly, the results suggest that providing general instruction regarding the kinds of conditions in which a decision rule does not apply caused them to fixate on the specific conditions mentioned and impaired their ability to identify other conditions in which the rule might not apply. The theoretical, methodological, and practical implications of the results are discussed.

Prediction of protein continuum secondary structure with probabilistic models based on NMR solved structures

Background: The structure of proteins may change as a result of the inherent flexibility of some protein regions. We develop and explore probabilistic machine learning methods for predicting a continuum secondary structure, i.e. assigning probabilities to the conformational states of a residue. We train our methods using data derived from high-quality NMR models. Results: Several probabilistic models not only successfully estimate the continuum secondary structure, but also provide a categorical output on par with models directly trained on categorical data. Importantly, models trained on the continuum secondary structure are also better than their categorical counterparts at identifying the conformational state for structurally ambivalent residues. Conclusion: Cascaded probabilistic neural networks trained on the continuum secondary structure exhibit better accuracy in structurally ambivalent regions of proteins, while sustaining an overall classification accuracy on par with standard, categorical prediction methods.

Physiologically based synthetic models of hepatic disposition

Current Physiologically based pharmacokinetic (PBPK) models are inductive. We present an additional, different approach that is based on the synthetic rather than the inductive approach to modeling and simulation. It relies on object-oriented programming A model of the referent system in its experimental context is synthesized by assembling objects that represent components such as molecules, cells, aspects of tissue architecture, catheters, etc. The single pass perfused rat liver has been well described in evaluating hepatic drug pharmacokinetics (PK) and is the system on which we focus. In silico experiments begin with administration of objects representing actual compounds. Data are collected in a manner analogous to that in the referent PK experiments. The synthetic modeling method allows for recognition and representation of discrete event and discrete time processes, as well as heterogeneity in organization, function, and spatial effects. An application is developed for sucrose and antipyrine, administered separately and together PBPK modeling has made extensive progress in characterizing abstracted PK properties but this has also been its limitation. Now, other important questions and possible extensions emerge. How are these PK properties and the observed behaviors generated? The inherent heuristic limitations of traditional models have hindered getting meaningful, detailed answers to such questions. Synthetic models of the type described here are specifically intended to help answer such questions. Analogous to wet-lab experimental models, they retain their applicability even when broken apart into sub-components. Having and applying this new class of models along with traditional PK modeling methods is expected to increase the productivity of pharmaceutical research at all levels that make use of modeling and simulation.

Promoting Learning for Sustainability: Principals' Perceptions of the Role of Outdoor and Environmental Education Centres

Outdoor and Environmental Education Centres provide programs that are designed to address a range of environmental education aims, and contribute broadly to student learning for sustainability. This paper examines the roles such Centres can play, and how they might contribute to the Australian Government’s initiative in relation to sustainable schools. Interviews with the principals of 23 such Centres in Queensland revealed three roles or models under which they operate: the destination model; the expert/advisor model; and the partnership model. Principals’ understandings of these roles are discussed and the factors that support or hinder their implementation are identified. It is concluded that while the provision of programs in the environment is still a vital role of outdoor and environmental education centres, these can also be seen as a point of entry to long-term partnerships with whole school communities.