This thesis studies the generalization ability of machine learning algorithms in a statistical setting. It focuses on the data-dependent analysis of the generalization performance of learning algorithms in order to make full use of the potential of the actual training sample from which these algorithms learn.¶ First, we propose an extension of the standard framework for the derivation of generalization bounds for algorithms taking their hypotheses from random classes of functions. This approach is motivated by the fact that the function produced by a learning algorithm based on a random sample of data depends on this sample and is therefore a random function. Such an approach avoids the detour of the worst-case uniform bounds as done in the standard approach. We show that the mechanism which allows one to obtain generalization bounds for random classes in our framework is based on a “small complexity” of certain random coordinate projections. We demonstrate how this notion of complexity relates to learnability and how one can explore geometric properties of these projections in order to derive estimates of rates of convergence and good confidence interval estimates for the expected risk. We then demonstrate the generality of our new approach by presenting a range of examples, among them the algorithm-dependent compression schemes and the data-dependent luckiness frameworks, which fall into our random subclass framework.¶ Second, we study in more detail generalization bounds for a specific algorithm which is of central importance in learning theory, namely the Empirical Risk Minimization algorithm (ERM). Recent results show that one can significantly improve the high-probability estimates for the convergence rates for empirical minimizers by a direct analysis of the ERM algorithm. These results are based on a new localized notion of complexity of subsets of hypothesis functions with identical expected errors and are therefore dependent on the underlying unknown distribution. We investigate the extent to which one can estimate these high-probability convergence rates in a data-dependent manner. We provide an algorithm which computes a data-dependent upper bound for the expected error of empirical minimizers in terms of the “complexity” of data-dependent local subsets. These subsets are sets of functions of empirical errors of a given range and can be determined based solely on empirical data. We then show that recent direct estimates, which are essentially sharp estimates on the high-probability convergence rate for the ERM algorithm, can not be recovered universally from empirical data.

This study investigates information literacy and scholarly communication within the processes of doctoral research and supervision at a distance. Both doctoral candidates and supervisors acknowledge information literacy deficiencies and it is suggested that disintermediation and the proliferation of information may contribute to those deficiencies. Further to this, the influence of pedagogic continuity—particularly in relation to the information seeking behaviour of candidates—is investigated, as is the concomitant aspect of how doctoral researchers practise scholarly communication. The well-documented and enduring problem for candidates of isolation from the research cultures of their universities is also scrutinised. The contentious issue of more formally involving librarians in the doctoral process is also considered, from the perspective of candidates and supervisors. Superimposed upon these topical and timely issues is the theoretical framework of adult learning theory, in particular the tenets of andragogy. The pedagogical-andragogical orientation of candidates and supervisors is established, demonstrating both the differences and similarities between candidates and supervisors, as are a number of independent variables, including a comparison of on-campus and off-campus candidates. Other independent variables include age, gender, DETYA (Department of Education, Training & Youth Affairs) category, enrolment type, stage of candidature, employment and status, type of doctorate, and English/non-English speaking background. The research methodology uses qualitative and quantitative techniques encompassing both data and methodological triangulation. The study uses two sets of questionnaires and a series of in-depth interviews with a sample of on-campus and off-campus doctoral candidates and supervisors from four Australian universities. Major findings include NESB candidates being more pedagogical than their ESB counterparts, and candidates and supervisors from the Sciences are more pedagogical than those from Arts, Humanities and Social Sciences, or Education. Candidates make a transition from a more dependent and pedagogically oriented approach to learning towards more of an independent and andragogical orientation over the duration of their candidature. However, over tune both on-campus and off-campus candidates become more isolated from the research cultures of their universities, and less happy with support received from their supervisors in relation to their literature reviews. Ill The study found large discrepancies in perception between the support supervisors believed they gave to candidates in relation to the literature review, and the support candidates believed they received. Information seeking becomes easier over time, but candidates face a dilemma with the proliferation of information, suggesting that disintermediation has exacerbated the challenges of evaluation and organisation of information. The concept of pedagogic continuity was recognised by supervisors and especially candidates, both negative and positive influences. The findings are critically analysed and synthesised using the metaphor of a scholarly 'Club' of which obtaining a doctorate is a rite of passage. Recommendations are made for changes in professional practice, and topics that may warrant further research are suggested.

This field study examines the comparative effectiveness of traditional and humanistic teaching methods. In part I the origins of humanistic teaching and their relationship to Gestalt therapy are considered. The influence of Humanistic education adult learning theory is also discussed. Part I concludes with an assessment of humanistic teaching. Part II: To measure the comparative effectiveness of the humanistic and traditional teaching methods, three groups of first year students in General English classess in a College of Technical and Further Education were selected. The effectiveness was measured in terms of differences in 1. Self-esteem. 2. Spelling Ability. 3. Attrition rates. 4 Academic Results. It can be concluded that there is: No difference in measures of self-esteem between groups, on the basis of the teaching method used. A large increase in the measure of self-esteem was observed in one of the groups that was taught with humanistic teaching strategies. Spelling measures showed no significant changes in any of the groups. Attrition rates were significantly reduced in groups that received humanistic teaching. Academic results of students completing the course showed no significant differences.

<p>Under 1990-talet implementerades i Sverige en metod för behandling av aggressivitethos barn och unga, som kallas Aggression Replacement Training (ART). Metoden ärutvecklad i USA av Arnold P. Goldstein och hans kollegor. Metoden har sin teoretiskagrund inom den sociala inlärningsteorin och bygger till stor del på behandlingsprinciperinspirerade av kognitiv beteendeterapi. ART som behandlingsmetod för aggressiva barnoch unga har fått stor spridning i Sverige och en av de kommuner som erbjuder ARTsom öppenvårdsinsats inom socialtjänsten är Västerås. Denna uppsats utgår från ettuppdrag från familjebehandlingen i Västerås Stad, som önskar belysa vilka eventuellaupplevda effekter deras variant av ART har gett. Förutom detta syftar uppsatsen ocksåtill att teoretiskt beskriva ART med avseende på teoretiskt ursprung och praktiskttillämpning. Studien, som har en kvalitativ ansats, utgår från en specifikbehandlingsgrupp om fem barn, vars föräldrar har intervjuats om vilka eventuellaeffekter de upplever att behandlingen gett. Resultaten är inte helt entydiga, då vissaföräldrar väldigt tydligt upplevt stora förbättringar när det gäller ilska och aggressivitethos sitt barn, medan andra föräldrar inte tycker sig se några större förändringar.Genomgående är dock föräldrarna nöjda med metoden och dess genomförande somsådant, men efterfrågar t.ex. intensivare och mer långvariga insatser för att ökamöjligheten för deras barn att ta till sig behandlingen fullt ut.</p> / <p>During the 1990’s a method of treatment for aggression and anger problems in childrenand youth was introduced in Sweden, witch is called Aggression Replacement Training(ART). ART was first developed in the United States, by Arnold P. Goldstein andcolleagues. The method has its theoretical foundation in the social learning theory and isinspired by the treatment principles found in cognitive behavioural therapy. ART as atreatment method for aggressive children and youth has been widely spread in Swedenand one of the cities that offers ART within their social services are Västerås. This essayis a mission from the unit of family treatment in Västerås Stad, who has an interest inelucidate possible treatment outcomes from their work with ART. In addition to that, thepurpose of this essay is to describe the theoretical foundations and practical appliancesof ART. This study, which has a qualitative approach, emanate from a specific treatmentgroup of five children, whose parents has been interviewed about what, if any, outcomethey experience that the treatment has resulted in. The results are not really univocal.Some of the parents experience very significant improvements in their child’sexpression of anger and aggressiveness, whereas some other parents don’t experienceany improvement at all. Consistently though, all of the parents are very pleased with themethod itself and its implementation, but some of them ask for more intensive and farreachinginterventions, to increase the possibilities for their children to fully ingest thetreatment.Key</p>

We present distribution independent bounds on the generalization misclassification performance of a family of kernel classifiers with margin. Support Vector Machine classifiers (SVM) stem out of this class of machines. The bounds are derived through computations of the $V_gamma$ dimension of a family of loss functions where the SVM one belongs to. Bounds that use functions of margin distributions (i.e. functions of the slack variables of SVM) are derived.

University of Macau / Faculty of Education

This dissertation is a multi-layered examination of the practice of voting, with a focus on the electoral turnout of immigrants. Chapter Two’s statistical analyses show that pre-migration cultural familiarity with democracy, formalized as levels of democratization in source countries, strongly shapes the likelihood of post-migration voting among Canadian immigrants. These origin effects, comparable in size to the best predictors of turnout that we have, exert a persistent influence – affecting turnout not only among the foreign-born, but also among the native-born second generation. Multilevel models demonstrate that the shifting source country composition of immigrant period-of-arrival cohorts provides an alternate explanation for what have previously been identified as generational, racial, and length of residence or ‘exposure’ effects among immigrant voters. This provides further evidence that voting is in most cases habitual, and raises questions about the acquisition, transmission, and reproduction of a voting practice. Chapter Three’s narratives of political development, gathered through in-depth, semi-structured interviews, confirm the importance of parental influence, and suggest that the ‘stickiness’ of practical capacities like voting may be the result of powerful processes of observational social learning. Providing a new twist on dominant models of political socialization, observation of parental voting appears to be the pivotal event in a path-dependent process of political learning, with acquisition of values and beliefs playing a supporting, rather than a leading role. Chapter Four reviews recent efforts among sociologists to amend action theory to make more room for habit, and these efforts are discussed in reference to contemporary research on turnout. I argue that these theoretical revisions still retain too sharp a focus on the cognitive aspects of practice. There is a lack of appreciation for the ways that action itself – our own previous actions and the actions of those close to us – can directly structure outcomes. Evidence from cognitive neuroscience is used to more precisely delineate habitual behaviour and thought. Where the intergenerational transmission of voting behaviour is concerned, culture is often coded directly into embodied practice. Efforts to encourage electoral participation should be built on a better understanding of voting’s substantial behavioural aspects.

This dissertation is a multi-layered examination of the practice of voting, with a focus on the electoral turnout of immigrants. Chapter Two’s statistical analyses show that pre-migration cultural familiarity with democracy, formalized as levels of democratization in source countries, strongly shapes the likelihood of post-migration voting among Canadian immigrants. These origin effects, comparable in size to the best predictors of turnout that we have, exert a persistent influence – affecting turnout not only among the foreign-born, but also among the native-born second generation. Multilevel models demonstrate that the shifting source country composition of immigrant period-of-arrival cohorts provides an alternate explanation for what have previously been identified as generational, racial, and length of residence or ‘exposure’ effects among immigrant voters. This provides further evidence that voting is in most cases habitual, and raises questions about the acquisition, transmission, and reproduction of a voting practice. Chapter Three’s narratives of political development, gathered through in-depth, semi-structured interviews, confirm the importance of parental influence, and suggest that the ‘stickiness’ of practical capacities like voting may be the result of powerful processes of observational social learning. Providing a new twist on dominant models of political socialization, observation of parental voting appears to be the pivotal event in a path-dependent process of political learning, with acquisition of values and beliefs playing a supporting, rather than a leading role. Chapter Four reviews recent efforts among sociologists to amend action theory to make more room for habit, and these efforts are discussed in reference to contemporary research on turnout. I argue that these theoretical revisions still retain too sharp a focus on the cognitive aspects of practice. There is a lack of appreciation for the ways that action itself – our own previous actions and the actions of those close to us – can directly structure outcomes. Evidence from cognitive neuroscience is used to more precisely delineate habitual behaviour and thought. Where the intergenerational transmission of voting behaviour is concerned, culture is often coded directly into embodied practice. Efforts to encourage electoral participation should be built on a better understanding of voting’s substantial behavioural aspects.

Under 1990-talet implementerades i Sverige en metod för behandling av aggressivitethos barn och unga, som kallas Aggression Replacement Training (ART). Metoden ärutvecklad i USA av Arnold P. Goldstein och hans kollegor. Metoden har sin teoretiskagrund inom den sociala inlärningsteorin och bygger till stor del på behandlingsprinciperinspirerade av kognitiv beteendeterapi. ART som behandlingsmetod för aggressiva barnoch unga har fått stor spridning i Sverige och en av de kommuner som erbjuder ARTsom öppenvårdsinsats inom socialtjänsten är Västerås. Denna uppsats utgår från ettuppdrag från familjebehandlingen i Västerås Stad, som önskar belysa vilka eventuellaupplevda effekter deras variant av ART har gett. Förutom detta syftar uppsatsen ocksåtill att teoretiskt beskriva ART med avseende på teoretiskt ursprung och praktiskttillämpning. Studien, som har en kvalitativ ansats, utgår från en specifikbehandlingsgrupp om fem barn, vars föräldrar har intervjuats om vilka eventuellaeffekter de upplever att behandlingen gett. Resultaten är inte helt entydiga, då vissaföräldrar väldigt tydligt upplevt stora förbättringar när det gäller ilska och aggressivitethos sitt barn, medan andra föräldrar inte tycker sig se några större förändringar.Genomgående är dock föräldrarna nöjda med metoden och dess genomförande somsådant, men efterfrågar t.ex. intensivare och mer långvariga insatser för att ökamöjligheten för deras barn att ta till sig behandlingen fullt ut. / During the 1990’s a method of treatment for aggression and anger problems in childrenand youth was introduced in Sweden, witch is called Aggression Replacement Training(ART). ART was first developed in the United States, by Arnold P. Goldstein andcolleagues. The method has its theoretical foundation in the social learning theory and isinspired by the treatment principles found in cognitive behavioural therapy. ART as atreatment method for aggressive children and youth has been widely spread in Swedenand one of the cities that offers ART within their social services are Västerås. This essayis a mission from the unit of family treatment in Västerås Stad, who has an interest inelucidate possible treatment outcomes from their work with ART. In addition to that, thepurpose of this essay is to describe the theoretical foundations and practical appliancesof ART. This study, which has a qualitative approach, emanate from a specific treatmentgroup of five children, whose parents has been interviewed about what, if any, outcomethey experience that the treatment has resulted in. The results are not really univocal.Some of the parents experience very significant improvements in their child’sexpression of anger and aggressiveness, whereas some other parents don’t experienceany improvement at all. Consistently though, all of the parents are very pleased with themethod itself and its implementation, but some of them ask for more intensive and farreachinginterventions, to increase the possibilities for their children to fully ingest thetreatment.Key

A subfield of artificial intelligence, machine learning (ML), is concerned with the development of algorithms that allow computers to &ldquo / learn&rdquo / . ML is the process of training a system with large number of examples, extracting rules and finding patterns in order to make predictions on new data points (examples). The most common machine learning schemes are supervised, semi-supervised, unsupervised and reinforcement learning. These schemes apply to natural language processing, search engines, medical diagnosis, bioinformatics, detecting credit fraud, stock market analysis, classification of DNA sequences, speech and hand writing recognition in computer vision, to encounter just a few. In this thesis, we focus on Support Vector Machines (SVMs) which is one of the most powerful methods currently in machine learning. As a first motivation, we develop a model selection tool induced into SVM in order to solve a particular problem of computational biology which is prediction of eukaryotic pro-peptide cleavage site applied on the real data collected from NCBI data bank. Based on our biological example, a generalized model selection method is employed as a generalization for all kinds of learning problems. In ML algorithms, one of the crucial issues is the representation of the data. Discrete geometric structures and, especially, linear separability of the data play an important role in ML. If the data is not linearly separable, a kernel function transforms the nonlinear data into a higher-dimensional space in which the nonlinear data are linearly separable. As the data become heterogeneous and large-scale, single kernel methods become insufficient to classify nonlinear data. Convex combinations of kernels were developed to classify this kind of data [8]. Nevertheless, selection of the finite combinations of kernels are limited up to a finite choice. In order to overcome this discrepancy, we propose a novel method of &ldquo / infinite&rdquo / kernel combinations for learning problems with the help of infinite and semi-infinite programming regarding all elements in kernel space. This will provide to study variations of combinations of kernels when considering heterogeneous data in real-world applications. Combination of kernels can be done, e.g., along a homotopy parameter or a more specific parameter. Looking at all infinitesimally fine convex combinations of the kernels from the infinite kernel set, the margin is maximized subject to an infinite number of constraints with a compact index set and an additional (Riemann-Stieltjes) integral constraint due to the combinations. After a parametrization in the space of probability measures, it becomes semi-infinite. We analyze the regularity conditions which satisfy the Reduction Ansatz and discuss the type of distribution functions within the structure of the constraints and our bilevel optimization problem. Finally, we adapted well known numerical methods of semiinfinite programming to our new kernel machine. We improved the discretization method for our specific model and proposed two new algorithms. We proved the convergence of the numerical methods and we analyzed the conditions and assumptions of these convergence theorems such as optimality and convergence.