New Techniques in Deep Representation Learning

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

Transposon-free regions in mammalian genomes

Despite the presence of over 3 million transposons separated on average by similar to 500 bp, the human and mouse genomes each contain almost 1000 transposon-free regions (TFRs) over 10 kb in length. The majority of human TFRs correlate with orthologous TFRs in the mouse, despite the fact that most transposons are lineage specific. Many human TFRs also overlap with orthologous TFRs in the marsupial opossum, indicating that these regions have remained refractory to transposon insertion for long evolutionary periods. Over 90% of the bases covered by TFRs are noncoding, much of which is not highly conserved. Most TFRs are not associated with unusual nucleotide composition, but are significantly associated with genes encoding developmental regulators, suggesting that they represent extended regions of regulatory information that are largely unable to tolerate insertions, a conclusion difficult to reconcile with current conceptions of gene regulation.

Knowledge acquisition from L2 specialist texts

This research aims to investigate knowledge acquisition and concept formation in the domain of economics and business studies through a foreign language, English, from the very initial to the very final stage of development in the context of Higher Education in Turkey. It traces both the processes as well as the product of acquisition in order to provide a detailed picture of how knowledge acquisition occurs. It aims to explore ways in which the acquisition process can be facilitated and promoted while prospective students of the Department of Economics and Business Administration receive a language training programme, following the completion of which they will join their academic community which offers part of its courses through the English language. The study draws upon (some) theories of mental representation of knowledge, such as schema, frame and script. The concept of discourse community with its characteristics is investigated, enculturation of prospective students to acquire knowledge of their domain through L2 is explored, and the crucial role of the constructivist theory in relation to knowledge acquisition is highlighted. The present study was conducted through a process of enculturation taking place partly at the language centre of Çukurova University and partly at the target discourse community. The data utilised for initiating knowledge acquisition was obtained by establishing a corpus of economics and business texts, which the learners are expected to read during their academic courses utilising computerised technology. The method of think aloud protocols was used to analyse processes taking place in knowledge acquisition, while the product of what was acquired was investigated by means of written recall protocols. It has been discovered that knowledge acquisition operates on the basis of analogical and to a certain extent metaphorical reasoning. The evidence obtained from the think aloud protocols showed that neophytes were able to acquire fundamental concepts of their future domain by reaching the level of shared understanding with the members of their target community of the faculty. Diaries and questionnaire analyses demonstrated that enculturation facilitated learners' transition from the language centre into the target community. Analyses of the written recall protocols and examinations from the post-enculturation stage of the research showed that neophytes' academic performances in their target community were much higher than those of their non-enculturated counterparts. Processes learners go through and strategies they spontaneously make use of, especially while acquiring knowledge of a specific domain through L2 have so far remained unexplored research areas. The present research makes a potential contribution to the language and knowledge acquisition theories by examining closely and systematically the language and the strategies they employ in acquiring such knowledge. The research findings offer useful implications to English language teaching at language schools. Language teachers are provided with useful guidelines as to how they can provide prospective students of a particular academic community with an experience of acquiring fundamental concepts of their discipline before they become members of their target community.

Resource-efficient supply chains:a research framework, literature review and research agenda

Purpose: To define a research agenda for creating Resource-Efficient Supply Chains (RESC) by identifying and analysing their key characteristics as well as future research opportunities. Design/methodology/approach: We follow a systematic review method to analyse the literature and to understand RESC taking a substantive theory approach. Our approach is grounded in a specific domain, the agri-food sector, because it is an intensive user of an extensive range of resources. Findings: The review shows that literature has looked at the use of resources primarily from the environmental impact perspective. It shows a lack of understanding of the specific RESC characteristics, and concludes more research is needed on multi-disciplinary methods for resource use and impact analyses as well as assessment methods for resource sensitivity and responsiveness. There is a need to explore whether or not, and how, logistics/supply chain decisions will affect the overall configuration of future food supply chains in an era of resource scarcity and depletion and what the trade-offs will be. Research limitations/implications: The paper proposes an agenda for future research in the area of resource–efficient supply chain. The framework proposed along with the key characteristics identified for RESC can be applied to other sectors. Practical implications: Our research should facilitate further understanding of the implications and trade-offs of supply chain decisions taken on the use of resources by supply chain managers. Originality/value: The paper explores the interaction between supply chains and natural resources and also defines the key characteristics of RESC.

INFRAWEBS Semantic Web Service Development on the Base of Knowledge Management Layer

The paper gives an overview about the ongoing FP6-IST INFRAWEBS project and describes the main layers and software components embedded in an application oriented realisation framework. An important part of INFRAWEBS is a Semantic Web Unit (SWU) – a collaboration platform and interoperable middleware for ontology-based handling and maintaining of SWS. The framework provides knowledge about a specific domain and relies on ontologies to structure and exchange this knowledge to semantic service development modules. INFRAWEBS Designer and Composer are sub-modules of SWU responsible for creating Semantic Web Services using Case-Based Reasoning approach. The Service Access Middleware (SAM) is responsible for building up the communication channels between users and various other modules. It serves as a generic middleware for deployment of Semantic Web Services. This software toolset provides a development framework for creating and maintaining the full-life-cycle of Semantic Web Services with specific application support.

Towards next generation vertical search engines

As the Web evolves unexpectedly fast, information grows explosively. Useful resources become more and more difficult to find because of their dynamic and unstructured characteristics. A vertical search engine is designed and implemented towards a specific domain. Instead of processing the giant volume of miscellaneous information distributed in the Web, a vertical search engine targets at identifying relevant information in specific domains or topics and eventually provides users with up-to-date information, highly focused insights and actionable knowledge representation. As the mobile device gets more popular, the nature of the search is changing. So, acquiring information on a mobile device poses unique requirements on traditional search engines, which will potentially change every feature they used to have. To summarize, users are strongly expecting search engines that can satisfy their individual information needs, adapt their current situation, and present highly personalized search results. ^ In my research, the next generation vertical search engine means to utilize and enrich existing domain information to close the loop of vertical search engine's system that mutually facilitate knowledge discovering, actionable information extraction, and user interests modeling and recommendation. I investigate three problems in which domain taxonomy plays an important role, including taxonomy generation using a vertical search engine, actionable information extraction based on domain taxonomy, and the use of ensemble taxonomy to catch user's interests. As the fundamental theory, ultra-metric, dendrogram, and hierarchical clustering are intensively discussed. Methods on taxonomy generation using my research on hierarchical clustering are developed. The related vertical search engine techniques are practically used in Disaster Management Domain. Especially, three disaster information management systems are developed and represented as real use cases of my research work.^

Immunity and Arginine Deprivation in Alzheimer's Disease

The pathogenesis of Alzheimer’s disease (AD) is a critical unsolved question, and while recent studies have demonstrated a strong association between altered brain immune responses and disease progression, the mechanistic cause of neuronal dysfunction and death is unknown. We have previously described the unique CVN-AD mouse model of AD, in which immune-mediated nitric oxide is lowered to mimic human levels, resulting in a mouse model that demonstrates the cardinal features of AD, including amyloid deposition, hyperphosphorylated and aggregated tau, behavioral changes and age-dependent hippocampal neuronal loss. Using this mouse model, we studied longitudinal changes in brain immunity in relation to neuronal loss and, contrary to the predominant view that AD pathology is driven by pro-inflammatory factors, we find that the pathology in CVN-AD mice is driven by local immune suppression. Areas of hippocampal neuronal death are associated with the presence of immunosuppressive CD11c+ microglia and extracellular arginase, resulting in arginine catabolism and reduced levels of total brain arginine. Pharmacologic disruption of the arginine utilization pathway by an inhibitor of arginase and ornithine decarboxylase protected the mice from AD-like pathology and significantly decreased CD11c expression. Our findings strongly implicate local immune-mediated amino acid catabolism as a novel and potentially critical mechanism mediating the age-dependent and regional loss of neurons in humans with AD.

There is a large interest in identifying, lineage tracing, and determining the physiologic roles of monophagocytes in Alzheimer’s disease. While Cx3cr1 knock-in fluorescent reporting and Cre expressing mice have been critical for studying neuroimmunology, mice that are homozygous null or hemizygous for CX3CR1 have perturbed neural development and immune responses. There is, therefore, a need for similar tools in which mice are CX3CR1+/+. Here, we describe a mouse where Cre is driven by the Cx3cr1 promoter on a bacterial artificial chromosome (BAC) transgene (Cx3cr1-CreBT) and the Cx3cr1 locus is unperturbed. Similarly to Cx3cr1-Cre knock-in mice, these mice express Cre in Ly6C-, but not Ly6C+, monocytes and tissue macrophages, including microglia. These mice represent a novel tool that maintains the Cx3cr1 locus while allowing for selective gene targeting in monocytes and tissue macrophages.

The study of immunity in Alzheimer’s requires the ability to identify and quantify specific immune cell subsets by flow cytometry. While it is possible to identify lymphocyte subsets based on cell lineage-specific markers, the lack of such markers in brain myeloid cell subsets has prevented the study of monocytes, macrophages and dendritic cells. By improving on tissue homogenization, we present a comprehensive protocol for flow cytometric analysis, that allows for the identification of several cell types that have not been previously identified by flow cytometry. These cell types include F4/80hi macrophages, which may be meningeal macrophages, IA/IE+ macrophages, which may represent perivascular macrophages, and dendritic cells. The identification of these cell types now allows for their study by flow cytometry in homeostasis and disease.

Towards Next Generation Vertical Search Engines

As the Web evolves unexpectedly fast, information grows explosively. Useful resources become more and more difficult to find because of their dynamic and unstructured characteristics. A vertical search engine is designed and implemented towards a specific domain. Instead of processing the giant volume of miscellaneous information distributed in the Web, a vertical search engine targets at identifying relevant information in specific domains or topics and eventually provides users with up-to-date information, highly focused insights and actionable knowledge representation. As the mobile device gets more popular, the nature of the search is changing. So, acquiring information on a mobile device poses unique requirements on traditional search engines, which will potentially change every feature they used to have. To summarize, users are strongly expecting search engines that can satisfy their individual information needs, adapt their current situation, and present highly personalized search results. In my research, the next generation vertical search engine means to utilize and enrich existing domain information to close the loop of vertical search engine's system that mutually facilitate knowledge discovering, actionable information extraction, and user interests modeling and recommendation. I investigate three problems in which domain taxonomy plays an important role, including taxonomy generation using a vertical search engine, actionable information extraction based on domain taxonomy, and the use of ensemble taxonomy to catch user's interests. As the fundamental theory, ultra-metric, dendrogram, and hierarchical clustering are intensively discussed. Methods on taxonomy generation using my research on hierarchical clustering are developed. The related vertical search engine techniques are practically used in Disaster Management Domain. Especially, three disaster information management systems are developed and represented as real use cases of my research work.

Methods and tools for analysis and management of risks and regulatory compliance in the healthcare sector: the hospital at home – HaH

Changing or creating an organisation means creating a new process. Each process involves many risks that need to be identified and managed. The main risks considered here are procedural and legal risks. The former are related to the risks of errors that may occur during processes, while the latter are related to the compliance of processes with regulations. Managing the risks implies proposing changes to the processes that allow the desired result: an optimised process. In order to manage a company and optimise it in the best possible way, not only should the organisational aspect, risk management and legal compliance be taken into account, but it is important that they are all analysed simultaneously with the aim of finding the right balance that satisfies them all. This is the aim of this thesis, to provide methods and tools to balance these three characteristics, and to enable this type of optimisation, ICT support is used. This work isn’t a thesis in computer science or law, but rather an interdisciplinary thesis. Most of the work done so far is vertical and in a specific domain. The particularity and aim of this thesis is not to carry out an in-depth analysis of a particular aspect, but rather to combine several important aspects, normally analysed separately, which however have an impact and influence each other. In order to carry out this kind of interdisciplinary analysis, the knowledge base of both areas was involved and the combination and collaboration of different experts in the various fields was necessary. Although the methodology described is generic and can be applied to all sectors, the case study considered is a new type of healthcare service that allows patients in acute disease to be hospitalised to their home. This provide the possibility to perform experiments using real hospital database.

Investigation of candidate division TM7, a recently recognized major lineage of the domain bacteria with no known pure-culture representatives

A molecular approach was used to investigate a recently described candidate division of the domain Bacteria, TM7, currently known only from environmental 16S ribosomal DNA sequence data, A number of TM7-specific primers and probes were designed and evaluated. Fluorescence in situ hybridization (FISH) of a laboratory scale bioreactor using two independent TM7-specific probes revealed a conspicuous sheathed-filament morphotype, fortuitously enriched in the reactor. Morphologically, the filament matched the description of the Eikelboom morphotype 0041-0675 widely associated with bulking problems in activated-sludge wastewater treatment systems. Transmission electron microscopy of the bioreactor sludge demonstrated that the sheathed-filament morphotype had a typical gram-positive cell envelope ultrastructure. Therefore, TM7 is only the third bacterial lineage recognized to have gram-positive representatives. TM7-specific FISH analysis of two full-scale wastewater treatment plant sludges, including the one used to seed the laboratory scale reactor, indicated the presence of a number of morphotypes, including sheathed filaments. TM7-specific PCR clone libraries prepared from the two full-scale sludges yielded 23 novel TM7 sequences. Three subdivisions could be defined based on these data and publicly available sequences. Environmental sequence data and TM7-specific FISH analysis indicate that members of the TM7 division are present in a variety of terrestrial, aquatic, and clinical habitats. A highly atypical base substitution (Escherichia coli position 912; C to U) for bacterial 16S rRNAs was present in almost all TM7 sequences, suggesting that TM7 bacteria, like Archaea, may be streptomycin resistant at the ribosome level.

A novel approach to antigen-specific deletion of CTL with minimal cellular activation using alpha 3 domain mutants of MHC class I/peptide complex

In this study, we have compared the effector functions and fate of a number of human CTL clones in vitro or ex vivo following contact with variant peptides presented either on the cell surface or in a soluble multimeric format. In the presence of CD8 coreceptor binding, there is a good correlation between TCR signaling, killing of the targets, and Fast-mediated CTL apoptosis. Blocking CD8 binding using (alpha3 domain mutants of MHC class I results in much reduced signaling and reduced killing of the targets. Surprisingly, however, Fast expression is induced to a similar degree on these CTLs, and apoptosis of CTL is unaffected. The ability to divorce these events may allow the deletion of antigen-specific and pathological CTL populations without the deleterious effects induced by full CTL activation.

Domain-Specific Language for Coordination patterns

The integration and composition of software systems requires a good architectural design phase to speed up communications between (remote) components. However, during implementation phase, the code to coordinate such components often ends up mixed in the main business code. This leads to maintenance problems, raising the need for, on the one hand, separating the coordination code from the business code, and on the other hand, providing mechanisms for analysis and comprehension of the architectural decisions once made. In this context our aim is at developing a domain-specific language, CoordL, to describe typical coordination patterns. From our point of view, coordination patterns are abstractions, in a graph form, over the composition of coordination statements from the system code. These patterns would allow us to identify, by means of pattern-based graph search strategies, the code responsible for the coordination of the several components in a system. The recovering and separation of the architectural decisions for a better comprehension of the software is the main purpose of this pattern language

Child programming: an adequate domain specific language for programming specific robots

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

A domain specific language for domotic systems

To cope with modernity, the interesting of having a fully automated house has been increasing over the years, as technology evolves and as our lives become more stressful and overloaded. An automation system provides a way to simplify some daily tasks, allowing us to have more spare time to perform activities where we are really needed. There are some systems in this domain that try to implement these characteristics, but this kind of technology is at its early stages of evolution being that it is still far away of empowering the user with the desired control over a habitation. The reason is that the mentioned systems miss some important features such as adaptability, extension and evolution. These systems, developed from a bottom-up approach, are often tailored for programmers and domain experts, discarding most of the times the end users that remain with unfinished interfaces or products that they have difficulty to control. Moreover, complex behaviors are avoided, since they are extremely difficult to implement mostly due to the necessity of handling priorities, conflicts and device calibration. Besides, these solutions are only reachable at very high costs, yet they still have the limitation of being difficult to configure by non-technical people once in runtime operation. As a result, it is necessary to create a tool that allows the execution of several automated actions, with an interface that is easy to use but at the same time supports all the main features of this domain. It is also desirable that this tool is independent of the hardware so it can be reused, thus a Model Driven Development approach (MDD) is the ideal option, as it is a method that follows those principles. Since the automation domain has some very specific concepts, the use of models should be combined with a Domain Specific Language (DSL). With these two methods, it is possible to create a solution that is adapted to the end users, but also to domain experts and programmers due to the several levels of abstraction that can be added to diminish the complexity of use. The aim of this thesis is to design a Domain Specific Language (DSL) that uses the Model Driven Development approach (MDD), with the purpose of supporting Home Automation (HA) concepts. In this implementation, the development of simple and complex scenarios should be supported and will be one of the most important concerns. This DSL should also support other significant features in this domain, such as the ability to schedule tasks, which is something that is limited in the current existing solutions.

Learning domain-specific sentiment lexicons with applications to recommender systems

Search is now going beyond looking for factual information, and people wish to search for the opinions of others to help them in their own decision-making. Sentiment expressions or opinion expressions are used by users to express their opinion and embody important pieces of information, particularly in online commerce. The main problem that the present dissertation addresses is how to model text to find meaningful words that express a sentiment. In this context, I investigate the viability of automatically generating a sentiment lexicon for opinion retrieval and sentiment classification applications. For this research objective we propose to capture sentiment words that are derived from online users’ reviews. In this approach, we tackle a major challenge in sentiment analysis which is the detection of words that express subjective preference and domain-specific sentiment words such as jargon. To this aim we present a fully generative method that automatically learns a domain-specific lexicon and is fully independent of external sources. Sentiment lexicons can be applied in a broad set of applications, however popular recommendation algorithms have somehow been disconnected from sentiment analysis. Therefore, we present a study that explores the viability of applying sentiment analysis techniques to infer ratings in a recommendation algorithm. Furthermore, entities’ reputation is intrinsically associated with sentiment words that have a positive or negative relation with those entities. Hence, is provided a study that observes the viability of using a domain-specific lexicon to compute entities reputation. Finally, a recommendation system algorithm is improved with the use of sentiment-based ratings and entities reputation.