Dissertations / Theses on the topic 'Learning discrepancy'

The advent of Next Generation Sequencing (NGS), a little over a decade ago, has led to a vast and rapid increase in the generation of genomic data. The drastically reduced cost has in turn enabled powerful modifications that can be used to investigate not just genetic, but epigenetic, phenomena. Epigenetics refers to the study of mechanisms effecting gene expression other than the genetic code itself and thus, at the transcription level, incorporates DNA methylation, transcription factor binding and histone modifications amongst others. This thesis outlines and tackles two major challenges in the computational analysis of such data using techniques from machine learning. Firstly, I address the problem of testing for differential methylation between groups of bisulfite sequencing data sets. DNA methylation plays an important role in genomic imprinting, X-chromosome inactivation and the repression of repetitive elements, as well as being implicated in numerous diseases, such as cancer. Bisulfite sequencing provides single nucleotide resolution methylation data at the whole genome scale, but a sensitive analysis of such data is difficult. I propose a solution that uses a powerful kernel-based machine learning technique, the Maximum Mean Discrepancy, to leverage well-characterised spatial correlations in DNA methylation, and adapt the method for this particular use. I use this tailored method to analyse a novel data set from a study of ageing in three different tissues in the mouse. This study motivates further modifications to the method and highlights the utility of the underlying measure as an exploratory tool for methylation analysis. Secondly, I address the problem of predictive and explanatory modelling of chromatin immunoprecipitation sequencing data (ChIP-Seq). ChIP-Seq is typically used to assay the binding of a protein of interest, such as a transcription factor or histone, to the DNA, and as such is one of the most widely used sequencing assays. While peak callers are a powerful tool in identifying binding sites of sparse and clean ChIPSeq profiles, more broad signals defy analysis in this framework. Instead, generative models that explain the data in terms of the underlying sequence can help uncover mechanisms that predicting binding or the lack thereof. I explore current problems with ChIP-Seq analysis, such as zero-inflation and the use of the control experiment, known as the input. I then devise a method for representing k-mers that enables the use of longer DNA sub-sequences within a flexible model development framework, such as generalised linear models, without heavy programming requirements. Finally, I use these insights to develop an appropriate Bayesian generative model that predicts ChIP-Seq count data in terms of the underlying DNA sequence, incorporating DNA methylation information where available, fitting the model with the Expectation-Maximization algorithm. The model is tested on simulated data and real data pertaining to the histone mark H3k27me3. This thesis therefore straddles the fields of bioinformatics and machine learning. Bioinformatics is both plagued and blessed by the plethora of different techniques available for gathering data and their continual innovations. Each technique presents a unique challenge, and hence out-of-the-box machine learning techniques have had little success in solving biological problems. While I have focused on NGS data, the methods developed in this thesis are likely to be applicable to future technologies, such as Third Generation Sequencing methods, and the lessons learned in their adaptation will be informative for the next wave of computational challenges.

Many studies have investigated problems with the ability achievement discrepancy (AAD) method of diagnosing specific learning disabilities (SLDs). The definition of an SLD includes the presence of a deficit in one or more cognitive processing systems. Researchers in other studies found that the AAD method overdiagnoses English language learners and students from low socioeconomic backgrounds, and underdiagnoses students with cognitive processing deficits. Although SLD diagnostic methods have been widely researched, much less information is available regarding SLD diagnostic methods that predict important student outcomes, such as high school completion. The General Abilities Index (GAI) is an SLD diagnostic method that can identify cognitive processing deficits. This study examined the relationships between cognitive processing deficits and the GAI method, high school completion status, performance on state standards assessments, and SLD eligibility. Using a multivariate, nonexperimental design, this study analyzed 149 datasets from records of students tested for an SLD between 1996 to 2013. A GLM analysis found that several types of cognitive processing deficits predicted math and writing performance on the state standards assessment and predicted not being diagnosed with an SLD, while the GAI method failed to predict any relationship with the dependent variables. Positive social changes from this study may include improved SLD diagnostic practices and improved educational interventions that target the cognitive processing deficits. Improved educational outcomes for SLD persons may reduce the high rates of unemployment, substance abuse, and incarceration experienced by the adult SLD population.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2016-02-03T12:50:53Z No. of bitstreams: 1 robertaelianegadelhaaleixo.pdf: 734128 bytes, checksum: b9114d54a5e9d76ddd7940b59e04cf31 (MD5)
Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2016-02-03T13:20:47Z (GMT) No. of bitstreams: 1 robertaelianegadelhaaleixo.pdf: 734128 bytes, checksum: b9114d54a5e9d76ddd7940b59e04cf31 (MD5)
Made available in DSpace on 2016-02-03T13:20:47Z (GMT). No. of bitstreams: 1 robertaelianegadelhaaleixo.pdf: 734128 bytes, checksum: b9114d54a5e9d76ddd7940b59e04cf31 (MD5) Previous issue date: 2014-05-14
A presente dissertação objetivou a elaboração de uma proposta de intervenção para minimizar a defasagem de aprendizagem de Matemática em uma Escola Estadual de Educação Profissional (EEEP) no estado do Ceará. A partir de um caso de gestão, foram investigadas as condições de trabalho com a disciplina na instituição educacional para a proposição de alternativas à superação do problema encontrado. Esse recorte se justificou pelo fato de a autora deste trabalho, no início da pesquisa, ter sido gestora da EEEP em análise e, por isso, verificado que a defasagem de aprendizagem, especialmente em Matemática, pode se configurar como um dos principais entraves à implementação dos cursos profissionalizantes na escola. A fim de obter informações para descrever e analisar o caso, a investigação teve como metodologia o uso de entrevistas com roteiros semiestruturados e pesquisa documental. Ao final da descrição do caso no capítulo 1, levantou-se como hipóteses dois os elementos centrais que influenciam na existência do problema: a organização e as responsabilidades do trabalho da equipe gestora e o papel da gestão escolar na formação e no auxílio à atuação docente. No capítulo 2, o problema foi analisado levando-se em consideração esses dois elementos. A análise dos dados foi feita a partir da perspectiva de alguns autores: Heloísa Lück, Henry Mintzberg, Thelma Polon, José Carlos Libâneo, Márcia Lima, Ana Maria Falsarella, Sérgio Lorenzato, Plínio Moreira e Fernando Almeida. Desse modo, no capítulo 3, apresentou-se uma proposta de intervenção que consiste em ações para redefinir as atribuições da equipe gestora e organizar o seu trabalho, a fim de que a gestão possa atuar na formação e no auxílio à atuação docente, com foco no professor de Matemática. Dessa forma, nos limites desta investigação, proposições foram consideradas como uma tentativa de contribuir para superar a defasagem de aprendizagem em Matemática na escola pesquisada.
The present dissertation aimed to elaborate an intervention proposal to minimize the learning discrepancy in Mathematics at a State School of Professional Education (EEEP, in Portuguese) in the state of Ceará. From a management case, we investigated the work conditions with the discipline at the institution in order to propose alternatives to overcome the problem. The selection of the school is justified given that the author was, at the beginning of the research, the principal of the EEEP being analyzed and, therefore, in a position to verify that the learning discrepancy, especially in Mathematics, may configure one of the main hurdles to the implementation of professionalizing courses at the school. In order to obtain information to describe and analyze the case, the investigation had as methodology the use of interviews with semi-structured scripts and documental research. By the end of the first chapter we came to the conclusion that there are two core elements influencing the problem: the organization and the responsibilities of the management team and the role of the school management in teacher training and assistance. On chapter 2, the issue was analyzed taking these two elements into consideration. Data analysis was conducted from the perspective of the following authors: Heloísa Lück (2008, 2009), Henry Mintzberg (2010), Thelma Polon (2011), José Carlos Libâneo (2008), Márcia Lima (2007), Ana Maria Falsarella (2004), Sérgio Lorenzato (2006), Plínio Moreira (2005) and Fernando Almeida (2011). Therefore, on chapter 3 we presented an intervention proposal which consists on actions to redefine the attributions of the management team and organize their work, so that the management may act in teacher training and assistance, focusing on Mathematics teachers. As such, within the restraints of this research, proposals were considered as an attempt to contribute to overcome the learning discrepancy in Mathematics at the studied school.

Specific Learning Disabilities (SLD) have historically been difficult to define and measure which has led to uncertainty and controversy. The current study explored the practices of identifying specific learning disabilities in Kentucky by surveying school psychologist practitioners in the state. Information was obtained about current practices with regard to RTI implementation and methods and data used for SLD identification as well as the roles that school psychologists take in the response to intervention (RTI) process. The sample consisted of 97 current or recently (within the past year) practicing school psychologists from 45 districts across the state. It was predicted that the use of RTI data for SLD identification would be associated with the length of time a district had been implementing RTI. The data did not support such a relationship. The majority of the districts represented by respondents were noted to be beyond an initial implementation of RTI practices. Responses to questions regarding the implementation of core features of RTI were grouped into High Implementation (HI; n = 45) and Low Implementation (LI; n = 41) groups. An independent samples t-test found a significant difference between the HI and LI groups for the quality of implementation. The HI group evidenced higher quality ratings than the LI. The use of RTI data as the most frequent method for SLD determination was noted for 30.9% of respondents as opposed to 0% prior to 2007. However, severe discrepancy was the most preferred method (59.3%) used for determining placement followed by RTI (28.4%) and a pattern of strengths and weaknesses (4.9%). Districts were also not likely to utilize non-preferred types of data if a student transferred into their district with that non-preferred data. Finally, the roles of school psychologists in the RTI process were explored. Great variability was found across practitioners with regard to the roles they actively have in the RTI process; however, practitioners in the HI group generally were more involved in the RTI process than those in the LI group. The findings are discussed with regard to the current national SLD identification practices and the limitations of the current findings.

Doutoramento em Matemática Aplicada à Economia e Gestão
This thesis may be broadly divided into 4 parts. In the first part, we do a literature review of the state of the art in misspecification in Macroeconomics, and what so far has been the contribution of a relatively new area of research called Uncertainty Quantification to the Macroeconomics subject. These reviews are essential to contextualize the contribution of this thesis in the furthering of research dedicated to correcting non-linear misspecifications, and to account for several other sources of uncertainty, when modelling from an economic perspective. In the next three parts, we give an example, using the same simple DSGE model from macroeconomic theory, of how researchers may quantify uncertainty in a State-Space Model using a discrepancy term with a Gaussian Process prior. The second part of the thesis, we used a full Gaussian Process (GP) prior on the discrepancy term. Our experiments showed that despite the heavy computational constraints of our full GP method, we still managed to obtain a very interesting forecasting performance with such a restricted sample size, when compared with similar uncorrected DSGE models, or corrected DSGE models using state of the art methods for time series, such as imposing a VAR on the observation error of the state-space model. In the third part of our work, we improved on the computational performance of our previous method, using what has been referred in the literature as Hilbert Reduced Rank GP. This method has close links to Functional Analysis, and the Spectral Theorem for Normal Operators, and Partial Differential Equations. It indeed improved the computational processing time, albeit just slightly, and was accompanied with a similarly slight decrease in the forecasting performance. The fourth part of our work delved into how our method would account for model uncertainty just prior, and during, the great financial crisis of 2007-2009. Our technique allowed us to capture the crisis, albeit at a reduced applicability possibly due to computational constraints. This latter part also was used to deepen the understanding of our model uncertainty quantification technique with a GP. Identifiability issues were also studied. One of our overall conclusions was that more research is needed until this uncertainty quantification technique may be used in as part of the toolbox of central bankers and researchers for forecasting economic fluctuations, specially regarding the computational performance of either method.
info:eu-repo/semantics/publishedVersion

Förväntan på att företag arbetar aktivt med att minska sin miljöpåverkan växer sig större för var dag. Med en miljöcertifiering kan företag visa sig ansvarsfulla inför sina intressenter och chanserna för en säkrad plats på framtidens marknad kan öka. Miljöcertifikat såsom ISO 14001-standarden inkluderar behovet av ständiga förbättringar av en organisations processer och miljöstrategiskt arbete. Begreppet ständiga förbättringar har sitt ursprung i Toyota Motors framgångsrika arbetssätt - kaizen - och innebär att allt kan göras lite bättre än i nuläget. På så sätt klingar det väl med miljöförbäattringsarbete eftersom vi fortfarande är på en resa mot det mest miljövänliga sättet att leva på. Två företag inom elinstallatörsbranchen har genom den webbaserade applikationen System C2 TM utmärkt sig i sitt arbete med ständiga förbättringar inom miljö. Eftersom en framgångsfaktor för ständiga förbättringar är utbildning kan det antas att dessa två företag har genomgått ett lärande som har lett till denna framgång, men vad som kännetecknat detta lärande är inte specificerat. Syftet med detta examensarbete är därför att undersöka vad en sådan utbildning bör innehålla för att inte bara leda till fler miljöförbättringar, utan även för att skapa förutsättningar för ett lärande inom ekologisk hållbar utveckling. För att generera kunskap kring detta genomfördes kvalitativa fokussamtal i två omgångar, med två  olika företag. Samtalen analyserades med hjälp av en tematisk analys. Resultaten visar att en utbildningsinsats som ämnar leda till ett aktivt miljöförbättringsarbete och ett kontinuerligt lärande ska sträcka sig över flera utbildningstillfällen. Tillfällena i insatsen bör ta vara på verksamhetens kunskap genom inklusion av medarbetare, knyta miljöfrågan till verksamheten, drivas av en engagerad ledning och ta vara på medarbetarnas personliga engagemang. Dessutom ställs krav på företagets kultur i form av att den ska vara öppen för ifrågasättande av rutiner samt drivas av en vilja att utvecklas. Resultaten visar även på likheter mellan konceptet ständiga förbättringar och lärande, och vidare föreslås djupare studier kring dessa likheter. Vidare forskning föreslås även i form av att testa de faktorer som detta examensarbete har resulterat i.
The expectations for a company to actively work on reducing their environmental impact are higher today than ever before. With an environmental certification, companies can show their responsibilities towards the environment to their stakeholders, and improve the chances to securing a position on the future market. An environmental certification, such as the standard ISO 14001, includes the need for continuous improvements to an organization’s processes and strategic environmental work. The term continuous improvements origins from Toyota Motors successful way of working - kaizen. It implies that everything can be madea little better than it is right now. This interpretation of continuous improvements goeswell together with environmental improvements in business since we are still on a journeytowards the most sustainable way of living. Two electrical contracting companies have through the web-based application System C2TM excelled in their work with continuous environmental improvements. Since a success factor to continuous improvements has proven to be education, one can assume that the two studied companies have a learning culture in their workplace that has led to this success. However, this learning culture is not specified. The purpose of this study is to examine what a training session in ecologically sustainable development should involve, not only lead to continuous improvements, but also to continuous learning. To generate knowledge about this, a qualitative study has been made. Two focus group conversations with each of the studied companies were made to gather data about their work with continuous environmental improvements. The data from the focus group conversations were analyzed through a thematic analysis. The results show that a training session that aims to lead to continuous improvements and learning within the organization consists of several sessions. The training sessions should make use of the knowledge within the organization, connect the environmental issue to the organization, involve an engaged leadership, and make use of the personal engagement amongst the coworkers. Apart from this, demands on the company culture are being made in order to create a learning organization. The culture should be open to challenge the existing routines and strive to develop the company to the better. The results from this study show that there are similarities between continuous improvements and learning. Suggested as further research is to examine the similarities between learning and continuous improvements. Another example for further research is to test the factors presented in this study, in a training session about ecologically sustainable development.

L’intérêt considérable est dédié à l’autorégulation chez les conducteurs âgés. Notre première étude compare l’autorégulation des conducteurs âgés à celle des jeunes. Nous supposons trouver des corrélations positives entre les évitements auto-déclarés des situations de conduite difficiles et le niveau auto-déclaré de déclins cognitifs chez les conducteurs âgés, non pas chez les jeunes. Or, le fait d’opérationnaliser l’autorégulation par ce biais ne permet ni d’examiner le niveau d’autorégulation comportemental, ni les facteurs susceptibles de conditionner un tel comportement. Deux études sont alors conçues de sorte à définir le modèle de l’apprentissage autorégulé le plus adapté pour l’examen des capacités d’autorégulation (étude 2), et à identifier les éléments susceptibles de modifier une autorégulation comportementale (étude 3). Dans l’étude 2, nous testons dans une expérience sur simulateur de conduite deux modèles métacognitifs, dont celui s’avérant plus compréhensif est maintenu comme paradigme expérimental pour la troisième étude. Cette dernière compare les patterns d’autorégulation de deux groupes de conducteurs âgés, dont un soumis à la menace du stéréotype. Les résultats indiquent que les conducteurs âgés basent leurs comportements d’autorégulation sur leurs propres expériences de la difficulté, et que le fait de vouloir les instruire à l’aide d’indices extrinsèques semblables à ceux du stéréotype, puisse mener plus à une réduction de leur capacités qu’à leur supposée amélioration
An increasing interest is actually devoted to the self-regulation in elderly drivers. Our first study compares the elderly drivers’ self-regulation to that of younger drivers. We suppose to observe in elderly drivers (but not in the younger ones) positive correlations between the self-declared avoidance of difficult driving situations and the self-declared decline of cognitive functioning. However, such an operationalization of self-regulation does not allow the study of neither the real behavioral self-regulation nor the factors conditioning such a behavior. Another two studies are then conceived in order to define both the self-regulated learning model that fits the best to the study of drivers’ self-regulatory patterns (study 2) and the factors that trigger such a behavioral self-regulation (study 3). In the second, driving-simulator study, we oppose the predictions of two metacognitive models, and this model that is better suited is withheld as the experimental paradigm for the third study. This latter compare the self-regulatory patterns of two groups of elderly drivers: one that is under stereotype threat and one that is not. Our results indicate that, first, elderly drivers presumably base their self-regulatory behavior on their sheer experience of the difficulty and, second, the efforts to enhance their self-regulation by extrinsic cues (such as those similar to the stereotype used in our third study) in fact hinder their self-regulatory abilities rather than enhance them

Cette thèse de doctorat traite de l'inférence variationnelle et de la robustesse en statistique et en machine learning. Plus précisément, elle se concentre sur les propriétés statistiques des approximations variationnelles et sur la conception d'algorithmes efficaces pour les calculer de manière séquentielle, et étudie les estimateurs basés sur le Maximum Mean Discrepancy comme règles d'apprentissage qui sont robustes à la mauvaise spécification du modèle.Ces dernières années, l'inférence variationnelle a été largement étudiée du point de vue computationnel, cependant, la littérature n'a accordé que peu d'attention à ses propriétés théoriques jusqu'à très récemment. Dans cette thèse, nous étudions la consistence des approximations variationnelles dans divers modèles statistiques et les conditions qui assurent leur consistence. En particulier, nous abordons le cas des modèles de mélange et des réseaux de neurones profonds. Nous justifions également d'un point de vue théorique l'utilisation de la stratégie de maximisation de l'ELBO, un critère numérique qui est largement utilisé dans la communauté VB pour la sélection de modèle et dont l'efficacité a déjà été confirmée en pratique. En outre, l'inférence Bayésienne offre un cadre d'apprentissage en ligne attrayant pour analyser des données séquentielles, et offre des garanties de généralisation qui restent valables même en cas de mauvaise spécification des modèles et en présence d'adversaires. Malheureusement, l'inférence Bayésienne exacte est rarement tractable en pratique et des méthodes d'approximation sont généralement employées, mais ces méthodes préservent-elles les propriétés de généralisation de l'inférence Bayésienne ? Dans cette thèse, nous montrons que c'est effectivement le cas pour certains algorithmes d'inférence variationnelle (VI). Nous proposons de nouveaux algorithmes tempérés en ligne et nous en déduisons des bornes de généralisation. Notre résultat théorique repose sur la convexité de l'objectif variationnel, mais nous soutenons que notre résultat devrait être plus général et présentons des preuves empiriques à l'appui. Notre travail donne des justifications théoriques en faveur des algorithmes en ligne qui s'appuient sur des méthodes Bayésiennes approchées.Une autre question d'intérêt majeur en statistique qui est abordée dans cette thèse est la conception d'une procédure d'estimation universelle. Cette question est d'un intérêt majeur, notamment parce qu'elle conduit à des estimateurs robustes, un thème d'actualité en statistique et en machine learning. Nous abordons le problème de l'estimation universelle en utilisant un estimateur de minimisation de distance basé sur la Maximum Mean Discrepancy. Nous montrons que l'estimateur est robuste à la fois à la dépendance et à la présence de valeurs aberrantes dans le jeu de données. Nous mettons également en évidence les liens qui peuvent exister avec les estimateurs de minimisation de distance utilisant la distance L2. Enfin, nous présentons une étude théorique de l'algorithme de descente de gradient stochastique utilisé pour calculer l'estimateur, et nous étayons nos conclusions par des simulations numériques. Nous proposons également une version Bayésienne de notre estimateur, que nous étudions à la fois d'un point de vue théorique et d'un point de vue computationnel
This PhD thesis deals with variational inference and robustness. More precisely, it focuses on the statistical properties of variational approximations and the design of efficient algorithms for computing them in an online fashion, and investigates Maximum Mean Discrepancy based estimators as learning rules that are robust to model misspecification.In recent years, variational inference has been extensively studied from the computational viewpoint, but only little attention has been put in the literature towards theoretical properties of variational approximations until very recently. In this thesis, we investigate the consistency of variational approximations in various statistical models and the conditions that ensure the consistency of variational approximations. In particular, we tackle the special case of mixture models and deep neural networks. We also justify in theory the use of the ELBO maximization strategy, a model selection criterion that is widely used in the Variational Bayes community and is known to work well in practice.Moreover, Bayesian inference provides an attractive online-learning framework to analyze sequential data, and offers generalization guarantees which hold even under model mismatch and with adversaries. Unfortunately, exact Bayesian inference is rarely feasible in practice and approximation methods are usually employed, but do such methods preserve the generalization properties of Bayesian inference? In this thesis, we show that this is indeed the case for some variational inference algorithms. We propose new online, tempered variational algorithms and derive their generalization bounds. Our theoretical result relies on the convexity of the variational objective, but we argue that our result should hold more generally and present empirical evidence in support of this. Our work presents theoretical justifications in favor of online algorithms that rely on approximate Bayesian methods. Another point that is addressed in this thesis is the design of a universal estimation procedure. This question is of major interest, in particular because it leads to robust estimators, a very hot topic in statistics and machine learning. We tackle the problem of universal estimation using a minimum distance estimator based on the Maximum Mean Discrepancy. We show that the estimator is robust to both dependence and to the presence of outliers in the dataset. We also highlight the connections that may exist with minimum distance estimators using L2-distance. Finally, we provide a theoretical study of the stochastic gradient descent algorithm used to compute the estimator, and we support our findings with numerical simulations. We also propose a Bayesian version of our estimator, that we study from both a theoretical and a computational points of view

L'apprentissage statistique cherche à modéliser un lien fonctionnel entre deux variables X et Y à partir d'un échantillon aléatoire de réalisations de (X,Y ). Lorsque la variable Y prend un nombre binaire de valeurs, l'apprentissage s'appelle la classification (ou discrimination en français) et apprendre le lien fonctionnel s'apparente à apprendre la frontière d'une variété dans l'espace de la variable X. Dans cette thèse, nous nous plaçons dans le contexte de l'apprentissage actif, i.e. nous supposons que l'échantillon d'apprentissage n'est plus aléatoire et que nous pouvons, par l'intermédiaire d'un oracle, générer les points sur lesquels l'apprentissage de la variété va s'effectuer. Dans le cas où la variable Y est continue (régression), des travaux précédents montrent que le critère de la faible discrépance pour générer les premiers points d'apprentissage est adéquat. Nous montrons, de manière surprenante, que ces résultats ne peuvent pas être transférés à la classification. Dans ce manuscrit, nous proposons alors le critère de la dispersion pour la classification. Ce critère étant difficile à mettre en pratique, nous proposons un nouvel algorithme pour générer un plan d'expérience à faible dispersion dans le carré unité. Après une première approximation de la variété, des approximations successives peuvent être réalisées afin d'affiner la connaissance de celle-ci. Deux méthodes d'échantillonnage sont alors envisageables : le " selective sampling " qui choisit les points à présenter à un oracle parmi un ensemble fini de candidats et l'" adaptative sampling " qui permet de choisir n'importe quels points de l'espace de la variable X. Le deuxième échantillonnage peut être vu comme un passage à la limite du premier. Néanmoins, en pratique, il n'est pas raisonnable d'utiliser cette méthode. Nous proposons alors un nouvel algorithme basé sur le critère de dispersion, menant de front exploitation et exploration, pour approximer une variété.

Regression is a central problem in statistics and machine learning with applications everywhere in science and technology. In probabilistic regression the relationship between a set of features and a real-valued target variable is modelled as a conditional probability distribution. There are cases where this distribution is very complex and not properly captured by simple approximations, such as assuming a normal distribution. This thesis investigates how conditional Generative Adversarial Networks (GANs) can be used to properly capture more complex conditional distributions. GANs have seen great success in generating complex high-dimensional data, but less work has been done on their use for regression problems. This thesis presents experiments to better understand how conditional GANs can be used in probabilistic regression. Different versions of GANs are extended to the conditional case and evaluated on synthetic and real datasets. It is shown that conditional GANs can learn to estimate a wide range of different distributions and be competitive with existing probabilistic regression models.

This study looks at two methods of identification for a specific learning disability in reading – Response to Intervention and Discrepancy Testing – and examines their efficacy with English Language Learners (ELL). A McNemar chi square analysis will be used to compare the outcomes of the two methods. Additionally, a multiway contingency table will be constructed and the association between English as a First Language (EFL)students and ELL students will be determined by using a conditional odds ratio. It is expected that both methods of learning disability testing will have difficulty distinguishing between a learning disability in reading and a language acquisition problem. The results will help prevent the misplacement of ELLs into Special Education.
text