In the last decade Support Vector Machines (SVMs) have emerged as an important learning technique for solving classification and regression problems in various fields, most notably in computational biology, finance and text categorization. This is due in part to built-in mechanisms to ensure good generalization which leads to accurate prediction, the use of kernel functions to model non-linear distributions, the ability to train relatively quickly on large data sets using novel mathematical optimization techniques and most significantly the possibility of theoretical analysis using computational learning theory. In this thesis, we discuss the theoretical basis and computational approaches to Support Vector Machines.

Machine learning has enjoyed astounding practical success in a wide range of applications in recent years-practical success that often hurries ahead of our theoretical understanding. The standard framework for machine learning theory assumes full supervision, that is, training data consists of correctly labeled iid examples from the same task that the learned classifier is supposed to be applied to. However, many practical applications successfully make use of the sheer abundance of data that is currently produced. Such data may not be labeled or may be collected from various sources. The focus of this thesis is to provide theoretical analysis of machine learning regimes where the learner is given such (possibly large amounts) of non-perfect training data. In particular, we investigate the benefits and limitations of learning with unlabeled data in semi-supervised learning and active learning as well as benefits and limitations of learning from data that has been generated by a task that is different from the target task (domain adaptation learning). For all three settings, we propose Probabilistic Lipschitzness to model the relatedness between the labels and the underlying domain space, and we discuss our suggested notion by comparing it to other common data assumptions.

No description available.

Thesis (Ph. D.)--University of California, San Diego, 2007. / Title from first page of PDF file (viewed January 14, 2008). Available via ProQuest Digital Dissertations. Vita. Includes bibliographical references. Faces are central stimuli in our everyday life, hence, face processing is a sophisticated and highly specialized cognitive ability, at which adults are experts and children are proficient. Unlike other visuospatial abilities, face perception develops very slowly, becoming adult-like only well into adolescence. Some performance disparities between children and adults may reflect differences in general cognitive abilities, such as attention and memory. Alternatively, performance differences can be attributed to specific cognitive strategies implemented during face processing by different age groups; or to the interaction between the improvement of general abilities throughout development and the refinement of face specific cognitive strategies. The intent of the current studies was to further assess the development of and relationship between cognitive strategies in face processing. Specifically, we investigated the behavior and neurocorrelates associated with holistic face processing in children (8- to 11-year-olds), adolescents, and adults, utilizing the composite face effect. The task requires participants to engage in both holistic and featural processing, but certain trials (aligned-same) elicit a visual illusion called the composite face effect (CFE, calculated as difference between misaligned-same and aligned-same trials), which is considered an index of holistic processing. All age groups (adults, adolescents, 8- to 9-year-olds, 10- to 11-year-olds) showed a CFE, suggesting reliance on holistic processing. Notably, about half of the 8- to 11-year-old children displayed adult-like behavior and adult-like CFE, suggesting their reliance on holistic processing. However, the other half of the children performed below-chance on aligned-same trials, displayed an extremely large CFE, and a significant difference between different trials, suggesting reliance on a featural strategy. Thus child age groups were regrouped according to their accuracy performance on the hardest condition (aligned-same trials) into high performing and low performing children. We hypothesize that the aligned-same trials were too taxing for low-performing children, thus they fell back into relying on simpler strategies such as a difference-detection featural strategy. In order to further investigate the CFE behavioral differences between age and performance groups, we completed an imaging study. For the fMRI study children were grouped by performance rather than age following the results of our behavioral study. Overall, our imaging results for the CFE, thus for holistic processing, resembled behavioral results in that adult and high performing child groups revealed a similar (but not identical) whole-brain pattern of activation, whereas the low performing child group showed a distinctive pattern of activation for the composite face effect. Adults and high performing children showed a pattern of activation spanning frontal, parietal, temporal, and occipital lobes. In contrast, low performing children revealed a pattern of activation that spanned frontal, temporal, cingulate, and cerebellar regions. Brain areas typically associated with face processing, such as the right fusiform gyrus and right inferior temporal gyrus did not reach significance for the low performing child group. These differences may be attributable to the use of different cognitive strategies. However, the extent of frontal and cingulate cortex activation in low performing children may also suggest that because the task was especially difficult for them, working memory resources were particularly taxed, thus affecting the neural network engaged. Importantly, not only were performance differences associated with distinct neurocorrelates (i.e., differing profiles for low performing children vs. high performing children and adults), but age differences also had an appreciable effect. In fact, high performing children did not significantly differ from adults in the behavioral CFE, but did show differences in the neural CFE.

Thesis (Ph. D.)--University of Washington, 2007. / Includes bibliographical references (p. 134-142).

Thesis (Ph. D.)--Georgia State University, 2007. / Joyce E. Many, committee chair; Mona W. Matthews, Dana L. Fox, Lori N. Elliott, committee members. Electronic text (190 p. : ill. (some col.)) : digital, PDF file. Description based on contents viewed June 19, 2008; title from file title page. Includes bibliographical references (p. 177-187).

Thesis (Ph. D.)--University of Missouri-Columbia, 2002. / Typescript. Vita. Includes bibliographical references (leaves 130-138). Also available on the Internet.

Thesis (Ph. D.)--University of Missouri-Columbia, 2002. / Typescript. Vita. Includes bibliographical references (leaves 130-138). Also available on the Internet.

Thesis (Ph.D.)--City University of Hong Kong, 2009. / "Submitted to Department of Mathematics in partial fulfillment of the requirements for the degree of Doctor of Philosophy." Includes bibliographical references (leaves [57]-62)

Thesis (Ph.D.)--Ohio State University, 2005. / Title from first page of PDF file. Document formatted into pages; contains xv, 201 p.; also includes graphics. Includes bibliographical references (p. 191-201). Available online via OhioLINK's ETD Center