Computational modelling of non-equilibrium condensing steam flows in low-pressure steam turbines

The steam turbines play a significant role in global power generation. Especially, research on low pressure (LP) steam turbine stages is of special importance for steam turbine man- ufactures, vendors, power plant owners and the scientific community due to their lower efficiency than the high pressure steam turbine stages. Because of condensation, the last stages of LP turbine experience irreversible thermodynamic losses, aerodynamic losses and erosion in turbine blades. Additionally, an LP steam turbine requires maintenance due to moisture generation, and therefore, it is also affecting on the turbine reliability. Therefore, the design of energy efficient LP steam turbines requires a comprehensive analysis of condensation phenomena and corresponding losses occurring in the steam tur- bine either by experiments or with numerical simulations. The aim of the present work is to apply computational fluid dynamics (CFD) to enhance the existing knowledge and understanding of condensing steam flows and loss mechanisms that occur due to the irre- versible heat and mass transfer during the condensation process in an LP steam turbine. Throughout this work, two commercial CFD codes were used to model non-equilibrium condensing steam flows. The Eulerian-Eulerian approach was utilised in which the mix- ture of vapour and liquid phases was solved by Reynolds-averaged Navier-Stokes equa- tions. The nucleation process was modelled with the classical nucleation theory, and two different droplet growth models were used to predict the droplet growth rate. The flow turbulence was solved by employing the standard k-ε and the shear stress transport k-ω turbulence models. Further, both models were modified and implemented in the CFD codes. The thermodynamic properties of vapour and liquid phases were evaluated with real gas models. In this thesis, various topics, namely the influence of real gas properties, turbulence mod- elling, unsteadiness and the blade trailing edge shape on wet-steam flows, are studied with different convergent-divergent nozzles, turbine stator cascade and 3D turbine stator-rotor stage. The simulated results of this study were evaluated and discussed together with the available experimental data in the literature. The grid independence study revealed that an adequate grid size is required to capture correct trends of condensation phenomena in LP turbine flows. The study shows that accurate real gas properties are important for the precise modelling of non-equilibrium condensing steam flows. The turbulence modelling revealed that the flow expansion and subsequently the rate of formation of liquid droplet nuclei and its growth process were affected by the turbulence modelling. The losses were rather sensitive to turbulence modelling as well. Based on the presented results, it could be observed that the correct computational prediction of wet-steam flows in the LP turbine requires the turbulence to be modelled accurately. The trailing edge shape of the LP turbine blades influenced the liquid droplet formulation, distribution and sizes, and loss generation. The study shows that the semicircular trailing edge shape predicted the smallest droplet sizes. The square trailing edge shape estimated greater losses. The analysis of steady and unsteady calculations of wet-steam flow exhibited that in unsteady simulations, the interaction of wakes in the rotor blade row affected the flow field. The flow unsteadiness influenced the nucleation and droplet growth processes due to the fluctuation in the Wilson point.

Seeing Turkish state formation processes: Mapping language and education census data

Language provides an interesting lens to look at state-building processes because of its cross-cutting nature. For example, in addition to its symbolic value and appeal, a national language has other roles in the process, including: (a) becoming the primary medium of communication which permits the nation to function efficiently in its political and economic life, (b) promoting social cohesion, allowing the nation to develop a common culture, and (c) forming a primordial basis for self-determination. Moreover, because of its cross-cutting nature, language interventions are rarely isolated activities. Languages are adopted by speakers, taking root in and spreading between communities because they are legitimated by legislation, and then reproduced through institutions like the education and military systems. Pádraig Ó’ Riagáin (1997) makes a case for this observing that “Language policy is formulated, implemented, and accomplishes its results within a complex interrelated set of economic, social, and political processes which include, inter alia, the operation of other non-language state policies” (p. 45). In the Turkish case, its foundational role in the formation of the Turkish nation-state but its linkages to human rights issues raises interesting issues about how socio-cultural practices become reproduced through institutional infrastructure formation. This dissertation is a country-level case study looking at Turkey’s nation-state building process through the lens of its language and education policy development processes with a focus on the early years of the Republic between 1927 and 1970. This project examines how different groups self-identified or were self-identified (as the case may be) in official Turkish statistical publications (e.g., the Turkish annual statistical yearbooks and the population censuses) during that time period when language and ethnicity data was made publicly available. The overarching questions this dissertation explores include: 1.What were the geo-political conditions surrounding the development and influencing the Turkish government’s language and education policies? 2.Are there any observable patterns in the geo-spatial distribution of language, literacy, and education participation rates over time? In what ways, are these traditionally linked variables (language, literacy, education participation) problematic? 3.What do changes in population identifiers, e.g., language and ethnicity, suggest about the government’s approach towards nation-state building through the construction of a civic Turkish identity and institution building? Archival secondary source data was digitized, aggregated by categories relevant to this project at national and provincial levels and over the course of time (primarily between 1927 and 2000). The data was then re-aggregated into values that could be longitudinally compared and then layered on aspatial administrative maps. This dissertation contributes to existing body of social policy literature by taking an interdisciplinary approach in looking at the larger socio-economic contexts in which language and education policies are produced.

Structured Approaches for Exploring Interpersonal Relationships in Natural Language Text

Human relationships have long been studied by scientists from domains like sociology, psychology, literature, etc. for understanding people's desires, goals, actions and expected behaviors. In this dissertation we study inter-personal relationships as expressed in natural language text. Modeling inter-personal relationships from text finds application in general natural language understanding, as well as real-world domains such as social networks, discussion forums, intelligent virtual agents, etc. We propose that the study of relationships should incorporate not only linguistic cues in text, but also the contexts in which these cues appear. Our investigations, backed by empirical evaluation, support this thesis, and demonstrate that the task benefits from using structured models that incorporate both types of information. We present such structured models to address the task of modeling the nature of relationships between any two given characters from a narrative. To begin with, we assume that relationships are of two types: cooperative and non-cooperative. We first describe an approach to jointly infer relationships between all characters in the narrative, and demonstrate how the task of characterizing the relationship between two characters can benefit from including information about their relationships with other characters in the narrative. We next formulate the relationship-modeling problem as a sequence prediction task to acknowledge the evolving nature of human relationships, and demonstrate the need to model the history of a relationship in predicting its evolution. Thereafter, we present a data-driven method to automatically discover various types of relationships such as familial, romantic, hostile, etc. Like before, we address the task of modeling evolving relationships but don't restrict ourselves to two types of relationships. We also demonstrate the need to incorporate not only local historical but also global context while solving this problem. Lastly, we demonstrate a practical application of modeling inter-personal relationships in the domain of online educational discussion forums. Such forums offer opportunities for its users to interact and form deeper relationships. With this view, we address the task of identifying initiation of such deeper relationships between a student and the instructor. Specifically, we analyze contents of the forums to automatically suggest threads to the instructors that require their intervention. By highlighting scenarios that need direct instructor-student interactions, we alleviate the need for the instructor to manually peruse all threads of the forum and also assist students who have limited avenues for communicating with instructors. We do this by incorporating the discourse structure of the thread through latent variables that abstractly represent contents of individual posts and model the flow of information in the thread. Such latent structured models that incorporate the linguistic cues without losing their context can be helpful in other related natural language understanding tasks as well. We demonstrate this by using the model for a very different task: identifying if a stated desire has been fulfilled by the end of a story.

Compulsory licensing of standard-essential patents and injunctive relief as abuse of dominant position in violation of Article 102 TFEU

When a dominant undertaking holding a standard-essential patent uses its exclusive right to the IP to seek injunctions against those wishing to produce either de jure or de facto standard compliant products, it creates a conflict between the exclusive right to the use of the IP on the one hand and the possible abuse of dominance due to the exclusionary conduct on the other. The aim of the thesis is to focus on the issues concerning abuse of dominance in violation of Article 102 TFEU when the holder of the standard-essential patent seeks an injunction against a would-be licensee. The thesis is mainly based on the most recent ECJ case law in Huawei and the Commission’s recent decisions in Samsung and Motorola. The case law in Europe prior to those decisions was mainly focused on the German case law from Orange Book Standard which provided IP holders great leverage due to the almost automatic granting of injunctions against infringers. The ECJ in Huawei set out the requirements for when a de jure standard-essential patent holder would not be violating Article 102 TFEU when seeking an injunction, requiring that negotiations in good faith must take place prior to the seeking of the injunction and that all offers must comply with FRAND terms, thus limiting the scope of case law derived from Orange Book Standard in Germany. The ECJ chose not to follow all of the reasoning the Commission had laid out in Samsung and Motorola which provided a more licensee-friendly approach on the matter, but rather chose a compromise between the IP holder friendly German case law and the Commission’s decisions. However, the ECJ did not disclose how FRAND terms themselves should be interpreted, but rather left it for the national courts to decide. Furthermore, the thesis strongly argues that Huawei did not change the fact that only vertically integrated IP holders who have made a FRAND declaration are subject to the terms laid out in Huawei, thus leaving non-practicing entities such as patent trolls and entities that have not made a FRAND declaration outside its scope. The resulting conclusion from the thesis is that while the ECJ in Huawei presented new exceptional circumstances for when an IP holder could be abusing its dominant position when it seeks an injunction, it still left many more questions answered, such as the meaning of FRAND and whether deception in giving a FRAND declaration is prohibited under Article 102 TFEU or not.

Garbage collection optimization for non uniform memory access architectures

Cache-coherent non uniform memory access (ccNUMA) architecture is a standard design pattern for contemporary multicore processors, and future generations of architectures are likely to be NUMA. NUMA architectures create new challenges for managed runtime systems. Memory-intensive applications use the system’s distributed memory banks to allocate data, and the automatic memory manager collects garbage left in these memory banks. The garbage collector may need to access remote memory banks, which entails access latency overhead and potential bandwidth saturation for the interconnection between memory banks. This dissertation makes five significant contributions to garbage collection on NUMA systems, with a case study implementation using the Hotspot Java Virtual Machine. It empirically studies data locality for a Stop-The-World garbage collector when tracing connected objects in NUMA heaps. First, it identifies a locality richness which exists naturally in connected objects that contain a root object and its reachable set— ‘rooted sub-graphs’. Second, this dissertation leverages the locality characteristic of rooted sub-graphs to develop a new NUMA-aware garbage collection mechanism. A garbage collector thread processes a local root and its reachable set, which is likely to have a large number of objects in the same NUMA node. Third, a garbage collector thread steals references from sibling threads that run on the same NUMA node to improve data locality. This research evaluates the new NUMA-aware garbage collector using seven benchmarks of an established real-world DaCapo benchmark suite. In addition, evaluation involves a widely used SPECjbb benchmark and Neo4J graph database Java benchmark, as well as an artificial benchmark. The results of the NUMA-aware garbage collector on a multi-hop NUMA architecture show an average of 15% performance improvement. Furthermore, this performance gain is shown to be as a result of an improved NUMA memory access in a ccNUMA system. Fourth, the existing Hotspot JVM adaptive policy for configuring the number of garbage collection threads is shown to be suboptimal for current NUMA machines. The policy uses outdated assumptions and it generates a constant thread count. In fact, the Hotspot JVM still uses this policy in the production version. This research shows that the optimal number of garbage collection threads is application-specific and configuring the optimal number of garbage collection threads yields better collection throughput than the default policy. Fifth, this dissertation designs and implements a runtime technique, which involves heuristics from dynamic collection behavior to calculate an optimal number of garbage collector threads for each collection cycle. The results show an average of 21% improvements to the garbage collection performance for DaCapo benchmarks.

The experience of teaching statistics to non-specialist students in Saudi universities

Undoubtedly, statistics has become one of the most important subjects in the modern world, where its applications are ubiquitous. The importance of statistics is not limited to statisticians, but also impacts upon non-statisticians who have to use statistics within their own disciplines. Several studies have indicated that most of the academic departments around the world have realized the importance of statistics to non-specialist students. Therefore, the number of students enrolled in statistics courses has vastly increased, coming from a variety of disciplines. Consequently, research within the scope of statistics education has been able to develop throughout the last few years. One important issue is how statistics is best taught to, and learned by, non-specialist students. This issue is controlled by several factors that affect the learning and teaching of statistics to non-specialist students, such as the use of technology, the role of the English language (especially for those whose first language is not English), the effectiveness of statistics teachers and their approach towards teaching statistics courses, students’ motivation to learn statistics and the relevance of statistics courses to the main subjects of non-specialist students. Several studies, focused on aspects of learning and teaching statistics, have been conducted in different countries around the world, particularly in Western countries. Conversely, the situation in Arab countries, especially in Saudi Arabia, is different; here, there is very little research in this scope, and what there is does not meet the needs of those countries towards the development of learning and teaching statistics to non-specialist students. This research was instituted in order to develop the field of statistics education. The purpose of this mixed methods study was to generate new insights into this subject by investigating how statistics courses are currently taught to non-specialist students in Saudi universities. Hence, this study will contribute towards filling the knowledge gap that exists in Saudi Arabia. This study used multiple data collection approaches, including questionnaire surveys from 1053 non-specialist students who had completed at least one statistics course in different colleges of the universities in Saudi Arabia. These surveys were followed up with qualitative data collected via semi-structured interviews with 16 teachers of statistics from colleges within all six universities where statistics is taught to non-specialist students in Saudi Arabia’s Eastern Region. The data from questionnaires included several types, so different techniques were used in analysis. Descriptive statistics were used to identify the demographic characteristics of the participants. The chi-square test was used to determine associations between variables. Based on the main issues that are raised from literature review, the questions (items scales) were grouped and five key groups of questions were obtained which are: 1) Effectiveness of Teachers; 2) English Language; 3) Relevance of Course; 4) Student Engagement; 5) Using Technology. Exploratory data analysis was used to explore these issues in more detail. Furthermore, with the existence of clustering in the data (students within departments within colleges, within universities), multilevel generalized linear models for dichotomous analysis have been used to clarify the effects of clustering at those levels. Factor analysis was conducted confirming the dimension reduction of variables (items scales). The data from teachers’ interviews were analysed on an individual basis. The responses were assigned to one of the eight themes that emerged from within the data: 1) the lack of students’ motivation to learn statistics; 2) students' participation; 3) students’ assessment; 4) the effective use of technology; 5) the level of previous mathematical and statistical skills of non-specialist students; 6) the English language ability of non-specialist students; 7) the need for extra time for teaching and learning statistics; and 8) the role of administrators. All the data from students and teachers indicated that the situation of learning and teaching statistics to non-specialist students in Saudi universities needs to be improved in order to meet the needs of those students. The findings of this study suggested a weakness in the use of statistical software applications in these courses. This study showed that there is lack of application of technology such as statistical software programs in these courses, which would allow non-specialist students to consolidate their knowledge. The results also indicated that English language is considered one of the main challenges in learning and teaching statistics, particularly in institutions where English is not used as the main language. Moreover, the weakness of mathematical skills of students is considered another major challenge. Additionally, the results indicated that there was a need to tailor statistics courses to the needs of non-specialist students based on their main subjects. The findings indicate that statistics teachers need to choose appropriate methods when teaching statistics courses.

Endowing NoSQL DBMS with SQL features through standard Call Level Interfaces

To store, update and retrieve data from database management systems (DBMS), software architects use tools, like call-level interfaces (CLI), which provide standard functionalities to interact with DBMS. However, the emerging of NoSQL paradigm, and particularly new NoSQL DBMS providers, lead to situations where some of the standard functionalities provided by CLI are not supported, very often due to their distance from the relational model or due to design constraints. As such, when a system architect needs to evolve, namely from a relational DBMS to a NoSQL DBMS, he must overcome the difficulties conveyed by the features not provided by NoSQL DBMS. Choosing the wrong NoSQL DBMS risks major issues with components requesting non-supported features. This paper focuses on how to deploy features that are not so commonly supported by NoSQL DBMS (like Stored Procedures, Transactions, Save Points and interactions with local memory structures) by implementing them in standard CLI.

Language Outcomes of the Play and Language for Autistic Youngsters (PLAY) Project Home Consultation model—An Extended Analysis

The current study is a post-hoc analysis of data from the original randomized control trial of the Play and Language for Autistic Youngsters (PLAY) Home Consultation program, a parent-mediated, DIR/Floortime based early intervention program for children with ASD (Solomon, Van Egeren, Mahone, Huber, & Zimmerman, 2014). We examined 22 children from the original RCT who received the PLAY program. Children were split into two groups (high and lower functioning) based on the ADOS module administered prior to intervention. Fifteen-minute parent-child video sessions were coded through the use of CHILDES transcription software. Child and maternal language, communicative behaviors, and communicative functions were assessed in the natural language samples both pre- and post-intervention. Results demonstrated significant improvements in both child and maternal behaviors following intervention. There was a significant increase in child verbal and non-verbal initiations and verbal responses in whole group analysis. Total number of utterances, word production, and grammatical complexity all significantly improved when viewed across the whole group of participants; however, lexical growth did not reach significance. Changes in child communicative function were especially noteworthy, and demonstrated a significant increase in social interaction and a significant decrease in non-interactive behaviors. Further, mothers demonstrated an increase in responsiveness to the child’s conversational bids, increased ability to follow the child’s lead, and a decrease in directiveness. When separated for analyses within groups, trends emerged for child and maternal variables, suggesting greater gains in use of communicative function in both high and low groups over changes in linguistic structure. Additional analysis also revealed a significant inverse relationship between maternal responsiveness and child non-interactive behaviors; as mothers became more responsive, children’s non-engagement was decreased. Such changes further suggest that changes in learned skills following PLAY parent training may result in improvements in child social interaction and language abilities.

Chronic low back pain among French healthcare workers and prognostic factors of return to work (RTW): a non-randomized controlled trial

International audience

The impact of geometric non-linearities on the fatigue analysis of trailing edge bond lines in wind turbine rotor blades

The accurate prediction of stress histories for the fatigue analysis is of utmost importance for the design process of wind turbine rotor blades. As detailed, transient, and geometrically non-linear three-dimensional finite element analyses are computationally weigh too expensive, it is commonly regarded sufficient to calculate the stresses with a geometrically linear analysis and superimpose different stress states in order to obtain the complete stress histories. In order to quantify the error from geometrically linear simulations for the calculation of stress histories and to verify the practical applicability of the superposition principal in fatigue analyses, this paper studies the influence of geometric non-linearity in the example of a trailing edge bond line, as this subcomponent suffers from high strains in span-wise direction. The blade under consideration is that of the IWES IWT-7.5-164 reference wind turbine. From turbine simulations the highest edgewise loading scenario from the fatigue load cases is used as the reference. A 3D finite element model of the blade is created and the bond line fatigue assessment is performed according to the GL certification guidelines in its 2010 edition, and in comparison to the latest DNV GL standard from end of 2015. The results show a significant difference between the geometrically linear and non-linear stress analyses when the bending moments are approximated via a corresponding external loading, especially in case of the 2010 GL certification guidelines. This finding emphasizes the demand to reconsider the application of the superposition principal in fatigue analyses of modern flexible rotor blades, where geometrical nonlinearities become significant. In addition, a new load application methodology is introduced that reduces the geometrically non-linear behaviour of the blade in the finite element analysis.

How teachers' perceptions affect the academic and language assessment of immigrant children

Recent research evidences inconsistencies in teachers' practice regarding skills assessment of L2 students. Scientific evidence supports that less experienced teachers have lower orientation toward multiple task-tests for non-native students. Research questions: Whether school teachers as having different teaching training and unequal teaching experience with non-native students perceive differently a four-skills scale. Purpose of the study: This study intends to analyse the importance degree between the four skills/tasks: reading, writing, speaking and listening, in the perspective of school teachers. Method: 77 teachers, aged 32-62, with (and without) experience in teaching and adapting materials for immigrant students, divided into six groups according to their scientific domain. Assessment tools included a scale for judgement of four academic tasks adapted from the original “Inventory of Undergraduate and Graduate Level: Reading, Writing, Speaking and Listening Tasks (Rosenfeld, Leung & Ottman, 2001). Main Findings: 1) different degrees of importance attributed by teachers on tasks that should be included in academic and language test for immigrant students; 2) perceptions of teachers are determined by predictors in this order: scientific domain, experience with multicultural classes and lower prediction from teaching service and age; 3) different results between american and portuguese samples answering the same questionnaire.

The big four skills: Teachers’ assumptions on measurement of non-native students cognition

The four-skills on tests for young native speakers commonly do not generate correlation incongruency concerning the cognitive strategies frequently reported. Considering the non-native speakers there are parse evidence to determine which tasks are important to assess properly the cognitive and academic language proficiency (Cummins, 1980; 2012). Research questions: It is of high probability that young students with origin in immigration significantly differ on their communication strategies and skills in a second language processing context (1); attached to this first assumption, it is supposed that teachers significantly differ depending on their scientific area and previous training (2). Purpose: This study intends to examine whether school teachers (K-12) as having different origin in scientific domain of teaching and training perceive differently an adapted four-skills scale, in European Portuguese. Research methods: 77 teachers of five areas scientific areas, mean of teaching year service = 32 (SD= 2,7), 57 males and 46 females (from basic and high school levels). Main findings: ANOVA (Effect size and Post-hoc Tukey tests) and linear regression analysis (stepwise method) revealed statistically significant differences among teachers of different areas, mainly between language teachers and science teachers. Language teachers perceive more accurately tasks in a multiple manner to the broad skills that require to be measured in non-native students. Conclusion: If teachers perceive differently the importance of the big-four tasks, there would be incongruence on skills measurement that teachers select for immigrant puppils. Non-balanced tasks and the teachers’ perceptions on evaluation and toward competence of students would likely determine limitations for academic and cognitive development of non-native students. Furthermore, results showed sufficient evidence to conclude that tasks are perceived differently by teachers toward importance of specific skills subareas. Reading skills are best considered compared to oral comphreension skills in non-native students.

Identification des grands utilisateurs de soins de santé chez les patients souffrant de la douleur chronique non cancéreuse et suivis en soins de première ligne

Contexte: La douleur chronique non cancéreuse (DCNC) génère des retombées économiques et sociétales importantes. L’identification des patients à risque élevé d’être de grands utilisateurs de soins de santé pourrait être d’une grande utilité; en améliorant leur prise en charge, il serait éventuellement possible de réduire leurs coûts de soins de santé. Objectif: Identifier les facteurs prédictifs bio-psycho-sociaux des grands utilisateurs de soins de santé chez les patients souffrant de DCNC et suivis en soins de première ligne. Méthodologie: Des patients souffrant d’une DCNC modérée à sévère depuis au moins six mois et bénéficiant une ordonnance valide d’un analgésique par un médecin de famille ont été recrutés dans des pharmacies communautaires du territoire du Réseau universitaire intégré de santé (RUIS), de l’Université de Montréal entre Mai 2009 et Janvier 2010. Ce dernier est composé des six régions suivantes : Mauricie et centre du Québec, Laval, Montréal, Laurentides, Lanaudière et Montérégie. Les caractéristiques bio-psycho-sociales des participants ont été documentées à l’aide d’un questionnaire écrit et d’une entrevue téléphonique au moment du recrutement. Les coûts directs de santé ont été estimés à partir des soins et des services de santé reçus au cours de l’année précédant et suivant le recrutement et identifiés à partir de la base de données de la Régie d’Assurance maladie du Québec, RAMQ (assureur publique de la province du Québec). Ces coûts incluaient ceux des hospitalisations reliées à la douleur, des visites à l’urgence, des soins ambulatoires et de la médication prescrite pour le traitement de la douleur et la gestion des effets secondaires des analgésiques. Les grands utilisateurs des soins de santé ont été définis comme étant ceux faisant partie du quartile le plus élevé de coûts directs annuels en soins de santé dans l’année suivant le recrutement. Des modèles de régression logistique multivariés et le critère d’information d’Akaike ont permis d’identifier les facteurs prédictifs des coûts directs élevés en soins de santé. Résultats: Le coût direct annuel médian en soins de santé chez les grands utilisateurs de soins de santé (63 patients) était de 7 627 CAD et de 1 554 CAD pour les utilisateurs réguliers (188 patients). Le modèle prédictif final du risque d’être un grand utilisateur de soins de santé incluait la douleur localisée au niveau des membres inférieurs (OR = 3,03; 95% CI: 1,20 - 7,65), la réduction de la capacité fonctionnelle liée à la douleur (OR = 1,24; 95% CI: 1,03 - 1,48) et les coûts directs en soins de santé dans l’année précédente (OR = 17,67; 95% CI: 7,90 - 39,48). Les variables «sexe», «comorbidité», «dépression» et «attitude envers la guérison médicale» étaient également retenues dans le modèle prédictif final. Conclusion: Les patients souffrant d’une DCNC au niveau des membres inférieurs et présentant une détérioration de la capacité fonctionnelle liée à la douleur comptent parmi ceux les plus susceptibles d’être de grands utilisateurs de soins et de services. Le coût direct en soins de santé dans l’année précédente était également un facteur prédictif important. Améliorer la prise en charge chez cette catégorie de patients pourrait influencer favorablement leur état de santé et par conséquent les coûts assumés par le système de santé.

Linear non-local diffusion problems in metric measure spaces

The aim of this paper is to provide a comprehensive study of some linear non-local diffusion problems in metric measure spaces. These include, for example, open subsets in ℝN, graphs, manifolds, multi-structures and some fractal sets. For this, we study regularity, compactness, positivity and the spectrum of the stationary non-local operator. We then study the solutions of linear evolution non-local diffusion problems, with emphasis on similarities and differences with the standard heat equation in smooth domains. In particular, we prove weak and strong maximum principles and describe the asymptotic behaviour using spectral methods.

SEQUENTIAL DECISIONS AND PREDICTIONS IN NATURAL LANGUAGE PROCESSING

Natural language processing has achieved great success in a wide range of ap- plications, producing both commercial language services and open-source language tools. However, most methods take a static or batch approach, assuming that the model has all information it needs and makes a one-time prediction. In this disser- tation, we study dynamic problems where the input comes in a sequence instead of all at once, and the output must be produced while the input is arriving. In these problems, predictions are often made based only on partial information. We see this dynamic setting in many real-time, interactive applications. These problems usually involve a trade-off between the amount of input received (cost) and the quality of the output prediction (accuracy). Therefore, the evaluation considers both objectives (e.g., plotting a Pareto curve). Our goal is to develop a formal understanding of sequential prediction and decision-making problems in natural language processing and to propose efficient solutions. Toward this end, we present meta-algorithms that take an existent batch model and produce a dynamic model to handle sequential inputs and outputs. Webuild our framework upon theories of Markov Decision Process (MDP), which allows learning to trade off competing objectives in a principled way. The main machine learning techniques we use are from imitation learning and reinforcement learning, and we advance current techniques to tackle problems arising in our settings. We evaluate our algorithm on a variety of applications, including dependency parsing, machine translation, and question answering. We show that our approach achieves a better cost-accuracy trade-off than the batch approach and heuristic-based decision- making approaches. We first propose a general framework for cost-sensitive prediction, where dif- ferent parts of the input come at different costs. We formulate a decision-making process that selects pieces of the input sequentially, and the selection is adaptive to each instance. Our approach is evaluated on both standard classification tasks and a structured prediction task (dependency parsing). We show that it achieves similar prediction quality to methods that use all input, while inducing a much smaller cost. Next, we extend the framework to problems where the input is revealed incremen- tally in a fixed order. We study two applications: simultaneous machine translation and quiz bowl (incremental text classification). We discuss challenges in this set- ting and show that adding domain knowledge eases the decision-making problem. A central theme throughout the chapters is an MDP formulation of a challenging problem with sequential input/output and trade-off decisions, accompanied by a learning algorithm that solves the MDP.