A Sparse Bayesian Approach for Joint Feature Selection and Classifier Learning

Peer-reviewed

Learning Preferences with Kernel-Based Methods

Learning of preference relations has recently received significant attention in machine learning community. It is closely related to the classification and regression analysis and can be reduced to these tasks. However, preference learning involves prediction of ordering of the data points rather than prediction of a single numerical value as in case of regression or a class label as in case of classification. Therefore, studying preference relations within a separate framework facilitates not only better theoretical understanding of the problem, but also motivates development of the efficient algorithms for the task. Preference learning has many applications in domains such as information retrieval, bioinformatics, natural language processing, etc. For example, algorithms that learn to rank are frequently used in search engines for ordering documents retrieved by the query. Preference learning methods have been also applied to collaborative filtering problems for predicting individual customer choices from the vast amount of user generated feedback. In this thesis we propose several algorithms for learning preference relations. These algorithms stem from well founded and robust class of regularized least-squares methods and have many attractive computational properties. In order to improve the performance of our methods, we introduce several non-linear kernel functions. Thus, contribution of this thesis is twofold: kernel functions for structured data that are used to take advantage of various non-vectorial data representations and the preference learning algorithms that are suitable for different tasks, namely efficient learning of preference relations, learning with large amount of training data, and semi-supervised preference learning. Proposed kernel-based algorithms and kernels are applied to the parse ranking task in natural language processing, document ranking in information retrieval, and remote homology detection in bioinformatics domain. Training of kernel-based ranking algorithms can be infeasible when the size of the training set is large. This problem is addressed by proposing a preference learning algorithm whose computation complexity scales linearly with the number of training data points. We also introduce sparse approximation of the algorithm that can be efficiently trained with large amount of data. For situations when small amount of labeled data but a large amount of unlabeled data is available, we propose a co-regularized preference learning algorithm. To conclude, the methods presented in this thesis address not only the problem of the efficient training of the algorithms but also fast regularization parameter selection, multiple output prediction, and cross-validation. Furthermore, proposed algorithms lead to notably better performance in many preference learning tasks considered.

First- and second-language phonological representations in the mental lexicon

Performance-based studies on the psychological nature of linguistic competence can conceal significant differences in the brain processes that underlie native versus nonnative knowledge of language. Here we report results from the brain activity of very proficient early bilinguals making a lexical decision task that illustrates this point. Two groups of SpanishCatalan early bilinguals (Spanish-dominant and Catalan-dominant) were asked to decide whether a given form was a Catalan word or not. The nonwords were based on real words, with one vowel changed. In the experimental stimuli, the vowel change involved a Catalan-specific contrast that previous research had shown to be difficult for Spanish natives to perceive. In the control stimuli, the vowel switch involved contrasts common to Spanish and Catalan. The results indicated that the groups of bilinguals did not differ in their behavioral and event-related brain potential measurements for the control stimuli; both groups made very few errors and showed a larger N400 component for control nonwords than for control words. However, significant differences were observed for the experimental stimuli across groups: Specifically, Spanish-dominant bilinguals showed great difficulty in rejecting experimental nonwords. Indeed, these participants not only showed very high error rates for these stimuli, but also did not show an error-related negativity effect in their erroneous nonword decisions. However, both groups of bilinguals showed a larger correctrelated negativity when making correct decisions about the experimental nonwords. The results suggest that although some aspects of a second language system may show a remarkable lack of plasticity (like the acquisition of some foreign contrasts), first-language representations seem to be more dynamic in their capacity of adapting and incorporating new information. &

Unca Don and Unca Scrooge’s Guide to History. Representations of the Past in the Disney Comics of Don Rosa

This thesis studies the various forms and layers of representations of the past that can be found in the Disney comics of Don Rosa. To stay true to the legacy of renowned comic book artist Carl Barks, Rosa has stopped time in the duck universe to the 1950’s: the decade when Barks created his most noted stories. There is a special feel of historicalness in Rosa’s duck stories, as his characters recall events that occurred in both Rosa’s own stories as well as Barks’. Rosa has shed new light to the past of the characters by writing and illustrating the history of Scrooge McDuck, one of the most beloved Disney characters. Rosa is also adamant that the historical facts used in his stories are always correct and based on thorough research. The methodological tools used in the analysis of the comics come from the fields of comic book studies, film theory, and history culture. Film and comics are recognized by many scholars as very similar media, which share elements that make them comparable in many ways. This thesis utilizes studies on historical film, narrative and genre, which provide valuable insight and comparisons for analysis. The thesis consists of three main chapters, the first of which deconstructs the duck universe in the stories in order to understand how the historicalness in them is created,and which outside elements might affect them, including the genre of Disney comics, publishers, and the legacy of Barks. The next chapter focuses on The Life and Times of Scrooge McDuck series, i.e. the stories which are located in the past. Such stories feature similar representations of history as for example Westerns. They also compress and alter history to meet the restrictions of the medium of comics. The last part focuses on the adventure stories which draw inspiration from for example mythology, and take the characters to strange and mystical, but yet historical worlds. Such treasure-hunting stories show similarity to the action-adventure genre in film and for example their stereotypical representations of foreign cultures. Finally, the chapter addresses the problematic of historical fiction and its capability to write history.

Language of the Future, Language of the PRC –Representations of Putonghua in South China Morning Post

As the national language of the PRC, the world's growing economic power and the sovereign of Hong Kong, Putonghua is a language with multiple facets of relevance for the current Special Administrative Region. This paper seeks to explore and explain different representations of Putonghua in Hong Kong's leading English-language newspaper South China Morning Post in articles published between January 2012 and February 2013. The representations are studied in the context of the different discourses in which they appear, some of which feature language(s) as a central theme and some more marginally. An overview is first presented of the scholarly research on the most important developments in Hong Kong's complex language scene from the beginnings of the colony until present day, with the aim of detecting developments and attitudes with potential relevance or parallels to the context of Putonghua today. The paper then reflects on the media and its role in producing and perpetuating discourses in the society, before turning to more practical considerations on Hong Kong's English and Chinese language media and the role of South China Morning Post in it. The methods used in analysing the discourses are those of discourse analysis, with textual analysis as its starting point, in which close attention is paid to linguistic forms as the concrete representations of meanings in a text. Particularly the immediate contexts of the appearances of the word “Putonghua” in the articles were studied carefully to detect vocabulary, grammar and semantical choices as signs of different discourses, potentially also revealing fundamental underlying assumptions and other “hidden meanings” in the text. Some of the most distinctive discourses in which different representations of Putonghua appeared were the Instrumental value for the individual (in which Putonghua was represented as a form of social capital); Othering of the mainlanders (in which Putonghua served as a concrete marker of distinction); Belonging to China (Putonghua as a symbol of unity); and Cultural distinctiveness of Hong Kong (Putonghua as a threat to Hong Kong's history and culture, as embodied in Cantonese). Some of these discourses were more prominent than others, and for example the discourse of Belonging to China was relatively rarely enacted in Hongkongers' voices. In general, the findings were not surprising in the light of the history, but showed a fair degree of consistency with what has been written earlier about the languages and attitudes towards them in Hong Kong. It has often been noted that Putonghua and its relation with Cantonese is a matter linked with the social identity of the colony and its citizens. While it appeared that there were no strict taboos in the representations of Putonghua in the societal context, the possibility of self-censorship cannot be ruled out as a factor toning down political discourses in the representations.

Representations of cultural encounters in the global business press – The importance of cross-cultural competence

This qualitative study has started from the interest to examine how the reality of crosscultural encounters is presented in the global business press. The research paper emphasizes different ways to classify culture and cross-cultural competency, both from the point of view of individuals and organizations. The analysis consists of public discourses, where cross-cultural realities are created through different persons, stories and contexts For data collection, a comprehensive database search was performed and 10 articles from the widely known worldwide business magazine The Financial Times were chosen as the data for the study paper. For the functions of addressing the research study questions, Thematic Content Analysis (TCA) and also Discourse Analysis (DA) are utilized, added with the continuous comparison method of grounded theory in the formation of the data.The academic references consist of literary works and articles presenting relevant concepts, creating a cross-cultural framework, and it is designed to assist the reader in the navigation through the topics of culture and cross-cultural competency. The repertoires were formed from the data and following, the first repertoire is contrast difference between home and target culture that the individual was able to discern. As a consequence of the first repertoire, the companies then offer cultural training to their employees to prepare them to situations of increasing levels of cultural variation. The third repertoire is increased awareness of other cultures, which is conveyed as a result of cultural training and contextual work experience. The fourth repertoire is globalization as an international business environment, where the people in the articles perform their job functions. It is stated in the conclusions that the representations emphasize Western values and personal traits in leadership.

Solving large sparse linear systems over the field GF(2)

Nowadays problem of solving sparse linear systems over the field GF(2) remain as a challenge. The popular approach is to improve existing methods such as the block Lanczos method (the Montgomery method) and the Wiedemann-Coppersmith method. Both these methods are considered in the thesis in details: there are their modifications and computational estimation for each process. It demonstrates the most complicated parts of these methods and gives the idea how to improve computations in software point of view. The research provides the implementation of accelerated binary matrix operations computer library which helps to make the progress steps in the Montgomery and in the Wiedemann-Coppersmith methods faster.

Representations of swinging London in 1960s British cinema : Blowup (1966), Smashing Time (1967), and Performance (1970)

This thesis explores the representation of Swinging London in three examples of 1960s British cinema: Blowup (Michelangelo Antonioni, 1966), Smashing Time (Desmond Davis, 1967) and Performance (Donald Cammell and Nicolas Roeg, 1970). It suggests that the films chronologically signify the evolution, commodification and dissolution of the Swinging London era. The thesis explores how the concept of Swinging London is both critiqued and perpetuated in each film through the use of visual tropes: the reconstruction of London as a cinematic space; the Pop photographer; the dolly; representations of music performance and fashion; the appropriation of signs and symbols associated with the visual culture of Swinging London. Using fashion, music performance, consumerism and cultural symbolism as visual narratives, each film also explores the construction of youth identity through the representation of manufactured and mediated images. Ultimately, these films reinforce Swinging London as a visual economy that circulates media images as commodities within a system of exchange. With this in view, the signs and symbols that comprise the visual culture of Swinging London are as central and significant to the cultural era as their material reality. While they attempt to destabilize prevailing representations of the era through the reproduction and exchange of such symbols, Blowup, Smashing Time, and Performance nevertheless contribute to the nostalgia for Swinging London in larger cultural memory.

Enabling and Measuring Complexity in Evolving Designs using Generative Representations for Artificial Architecture

As the complexity of evolutionary design problems grow, so too must the quality of solutions scale to that complexity. In this research, we develop a genetic programming system with individuals encoded as tree-based generative representations to address scalability. This system is capable of multi-objective evaluation using a ranked sum scoring strategy. We examine Hornby's features and measures of modularity, reuse and hierarchy in evolutionary design problems. Experiments are carried out, using the system to generate three-dimensional forms, and analyses of feature characteristics such as modularity, reuse and hierarchy were performed. This work expands on that of Hornby's, by examining a new and more difficult problem domain. The results from these experiments show that individuals encoded with those three features performed best overall. It is also seen, that the measures of complexity conform to the results of Hornby. Moving forward with only this best performing encoding, the system was applied to the generation of three-dimensional external building architecture. One objective considered was passive solar performance, in which the system was challenged with generating forms that optimize exposure to the Sun. The results from these and other experiments satisfied the requirements. The system was shown to scale well to the architectural problems studied.

A Study of Ordered Gene Problems Featuring DNA Error Correction and DNA Fragment Assembly with a Variety of Heuristics, Genetic Algorithm Variations, and Dynamic Representations

Ordered gene problems are a very common classification of optimization problems. Because of their popularity countless algorithms have been developed in an attempt to find high quality solutions to the problems. It is also common to see many different types of problems reduced to ordered gene style problems as there are many popular heuristics and metaheuristics for them due to their popularity. Multiple ordered gene problems are studied, namely, the travelling salesman problem, bin packing problem, and graph colouring problem. In addition, two bioinformatics problems not traditionally seen as ordered gene problems are studied: DNA error correction and DNA fragment assembly. These problems are studied with multiple variations and combinations of heuristics and metaheuristics with two distinct types or representations. The majority of the algorithms are built around the Recentering- Restarting Genetic Algorithm. The algorithm variations were successful on all problems studied, and particularly for the two bioinformatics problems. For DNA Error Correction multiple cases were found with 100% of the codes being corrected. The algorithm variations were also able to beat all other state-of-the-art DNA Fragment Assemblers on 13 out of 16 benchmark problem instances.

Non-negative matrix decomposition approaches to frequency domain analysis of music audio signals

On étudie l’application des algorithmes de décomposition matricielles tel que la Factorisation Matricielle Non-négative (FMN), aux représentations fréquentielles de signaux audio musicaux. Ces algorithmes, dirigés par une fonction d’erreur de reconstruction, apprennent un ensemble de fonctions de base et un ensemble de coef- ficients correspondants qui approximent le signal d’entrée. On compare l’utilisation de trois fonctions d’erreur de reconstruction quand la FMN est appliquée à des gammes monophoniques et harmonisées: moindre carré, divergence Kullback-Leibler, et une mesure de divergence dépendente de la phase, introduite récemment. Des nouvelles méthodes pour interpréter les décompositions résultantes sont présentées et sont comparées aux méthodes utilisées précédemment qui nécessitent des connaissances du domaine acoustique. Finalement, on analyse la capacité de généralisation des fonctions de bases apprises par rapport à trois paramètres musicaux: l’amplitude, la durée et le type d’instrument. Pour ce faire, on introduit deux algorithmes d’étiquetage des fonctions de bases qui performent mieux que l’approche précédente dans la majorité de nos tests, la tâche d’instrument avec audio monophonique étant la seule exception importante.

Mental illness in modern and contemporary theatre : An analysis of representations of mental illness in a selection of plays, accompanied by a new play about schizophrenia

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal