Many algorithms exist for learning how to act in a repeated game and most have theoretical guarantees associated with their behaviour. However, there are few experimental results about the empirical performance of these algorithms, which is important for any practical application of this work. Most of the empirical claims in the literature to date have been based on small experiments, and this has hampered the development of multiagent learning (MAL) algorithms with good performance properties. In order to rectify this problem, we have developed a suite of tools for running multiagent experiments called the Multiagent Learning Testbed (MALT). These tools are designed to facilitate running larger and more comprehensive experiments by removing the need to code one-off experimental apparatus. MALT also provides a number of public implementations of MAL algorithms—hopefully eliminating or reducing differences between algorithm implementations and increasing the reproducibility of results. Using this test-suite, we ran an experiment that is unprecedented in terms of the number of MAL algorithms used and the number of game instances generated. The results of this experiment were analyzed by using a variety of performance metrics—including reward, maxmin distance, regret, and several types of convergence. Our investigation also draws upon a number of empirical analysis methods. Through this analysis we found some surprising results: the most surprising observation was that a very simple algorithm—one that was intended for single-agent reinforcement problems and not multiagent learning— performed better empirically than more complicated and recent MAL algorithms.

Many algorithms exist for learning how to act in a repeated game and most have theoretical guarantees associated with their behaviour. However, there are few experimental results about the empirical performance of these algorithms, which is important for any practical application of this work. Most of the empirical claims in the literature to date have been based on small experiments, and this has hampered the development of multiagent learning (MAL) algorithms with good performance properties. In order to rectify this problem, we have developed a suite of tools for running multiagent experiments called the Multiagent Learning Testbed (MALT). These tools are designed to facilitate running larger and more comprehensive experiments by removing the need to code one-off experimental apparatus. MALT also provides a number of public implementations of MAL algorithms—hopefully eliminating or reducing differences between algorithm implementations and increasing the reproducibility of results. Using this test-suite, we ran an experiment that is unprecedented in terms of the number of MAL algorithms used and the number of game instances generated. The results of this experiment were analyzed by using a variety of performance metrics—including reward, maxmin distance, regret, and several types of convergence. Our investigation also draws upon a number of empirical analysis methods. Through this analysis we found some surprising results: the most surprising observation was that a very simple algorithm—one that was intended for single-agent reinforcement problems and not multiagent learning— performed better empirically than more complicated and recent MAL algorithms.

Many algorithms exist for learning how to act in a repeated game and most have theoretical guarantees associated with their behaviour. However, there are few experimental results about the empirical performance of these algorithms, which is important for any practical application of this work. Most of the empirical claims in the literature to date have been based on small experiments, and this has hampered the development of multiagent learning (MAL) algorithms with good performance properties. In order to rectify this problem, we have developed a suite of tools for running multiagent experiments called the Multiagent Learning Testbed (MALT). These tools are designed to facilitate running larger and more comprehensive experiments by removing the need to code one-off experimental apparatus. MALT also provides a number of public implementations of MAL algorithms—hopefully eliminating or reducing differences between algorithm implementations and increasing the reproducibility of results. Using this test-suite, we ran an experiment that is unprecedented in terms of the number of MAL algorithms used and the number of game instances generated. The results of this experiment were analyzed by using a variety of performance metrics—including reward, maxmin distance, regret, and several types of convergence. Our investigation also draws upon a number of empirical analysis methods. Through this analysis we found some surprising results: the most surprising observation was that a very simple algorithm—one that was intended for single-agent reinforcement problems and not multiagent learning— performed better empirically than more complicated and recent MAL algorithms. / Science, Faculty of / Computer Science, Department of / Graduate

Th.--Litt. angl.--Aix-Marseille 1, 1986.

No description available.

Made available in DSpace on 2019-03-29T23:40:30Z (GMT). No. of bitstreams: 0 Previous issue date: 2012-08-16 / The objective of this research is to investigate the techno-scientific discourse and its effects on the subject from the notion of rest. Supported by the reading of the iscontents in modernity in Freud and Lacan, this work turned to the investigation of human discontents nowadays in critically analyzing the techno-scientific efforts to produce a technical resolution to the discomfort of man and his culture. For this purpose, the rest was used as a reference which expresses the radicalism of the malaise inherent in the subject because he inhabits the realm of language, inserted in the culture. The relationship between technoscience and the rest was covered in three dimensions: The first dimension introduces the rest and its importance for understanding the constitution of the subject and his encounter with the inescapable anguish. It was highlighted the impossible nature of drive satisfaction as a matter of structure of the subject. The second dimension shows the rest presenting the impossible of the drive in the field of culture, the social bond. The theory of Lacan's discourse was presented as a basis for presentation and analysis of techno-scientific discourse and how it operates with the rest in social discourse. The third dimension introduces the central hypothesis of this study when it questions technoscience as a discourse and its subjective traps to the subjetc and his culture. It was stated that technoscience suggests that the impossible is within reach of the subject disrupting his fantasy relationship with the Knowledge and Truth, and that the subjective effect is the fascination for the rest, the index of the impossible to the drive s satisfaction. It was concluded that this fascination is a sideration so that the technoscientific discourse focuses on the narcissistic disposition of the I offering to the subject apprehend the rest and eliminate the discomfort inherent to the subject because he is trespassed by language. Thus, the object a as a rest ceases to be a factor of cause for the desire to become a center of fascination in a constant drive to the death instinct. As a result, the new man that can emerge is someone constantly visited by this residual part, that by resisting to its apprehension generates a effect of fascination launching the subject in a position marked by melancholy. Key-words: psychoanalysis, rest, discomfort, technoscience, discourse. / O objetivo desta pesquisa é investigar o discurso tecnocientífico e seus efeitos sobre o sujeito a partir da noção de resto. Apoiado na leitura do mal-estar na Modernidade em Freud e Lacan, este trabalho voltou-se para a investigação do mal-estar na atualidade analisando criticamente os esforços tecnocientíficos em produzir uma resolução técnica ao mal-estar do homem e de sua cultura. Para tanto, utilizou-se o resto como uma referência que expressa a radicalidade do mal-estar inerente ao sujeito por habitar o campo da linguagem, inserido na cultura. A relação entre tecnociência e resto foi abordada em três dimensões: A primeira dimensão introduz o resto e sua importância para compreensão da constituição do sujeito e seu inescapável encontro com a angústia. Foi destacado o caráter impossível da satisfação pulsional como fato de estrutura do sujeito. A segunda dimensão apresenta o resto evidenciando o impossível da pulsão no campo da cultura, ou seja, no laço social. A teoria dos discursos de Lacan foi apresentada como fundamento para a apresentação e análise do discurso tecnocientífico e o modo como ele opera com o resto no discurso social. A terceira dimensão lança a hipótese central deste estudo ao questionar a tecnociencia como discurso e suas armadilhas subjetivas para o sujeito e sua cultura. Afirmou-se que a tecnociencia sugere que o impossível encontra-se ao alcance do sujeito desarticulando sua relação fantasmática com o Saber e a Verdade e que, o efeito subjetivo é a fascinação pelo resto, índice do impossível da satisfação pulsional. Concluiu-se que tal fascinação é da ordem de uma sideração de modo que o discurso tecnocientífico incide sobre a disposição narcísica do Eu oferecendo ao sujeito apreender o resto e eliminar o mal-estar inerente à sua condição de ser da linguagem. Dessa forma, o objeto a como resto deixa então de ser fator de causação do desejo para se tornar pólo de fascinação num constante empuxo à pulsão de morte. Como resultado dessa articulação, o novo homem que pode emergir é alguém visitado constantemente por esta parte residual que, em não se deixar apreender gera uma verdadeira fascinação lançando o sujeito numa posição melancolizada. Palavras-chave: psicanálise, resto, mal-estar, tecnociência, discurso.

Made available in DSpace on 2019-04-05T23:06:46Z (GMT). No. of bitstreams: 0 Previous issue date: 2005-10-10 / Nowadays, a new conception of man has been promoted by the scientifictechnological discourse, bringing different ways to people establish their relationships with themselves and with the others. Thus, we re having the chance to observe the development of new ways to experience happiness, pleasure and suffering. That is, new ways of living. Therefore, it was this man and his vicissitudes that mobilized us to do this research. In order to argue the focused problem, we took psychoanalysis as our reference theory in view its narrow entailing between subject and culture. From there, we chose the following questionings as the guidelines of this work: considering the culture while a field of interdictions, how can we think the contemporary man in his capacity to undo limits or, simply, to deny its existence? And, from the previous question, how the happiness is experienced currently, in view the various possibilities of satisfaction offered to us? The passage we used to argue such questions was: initially, we detach the freudian conception of man through the restrictions he needs to be submitted, in view of his cultured constitution; after that, we comment how the biological knowledge, one of the biggest representative of the current scientific discourse, has been inserted in the social scope, reproducing the assertive that limits can, and must, be exceeded through the various possibilities of body manipulation; later, we made the characterization of the contemporary man through his search for permanent satisfactions, having the body as his preferential target of pleasures, leading him on ways that assure a position of excess; finally, having determined the happiness in Freud as something related to the proper choices of the subject in his relation with the world, we establish the differentiated ways followed by the contemporary man in his high expectations in respect of the happy being. Key Words: Scientific-technological discourse, psychoanalysis, happiness, moderation and excess. / O discurso da ciência e da tecnologia tem promovido, nos dias de hoje, uma nova concepção de homem que traz em si formas diferentes de relação consigo mesmo e com os outros. Estamos, pois, tendo a oportunidade de observar a construção de novas maneiras de se experimentar a felicidade, o prazer e o sofrimento. Em última instância, novas maneiras de se viver. Foi, portanto, esse homem e suas vicissitudes que nos mobilizaram a fazer esta pesquisa. A fim de discutirmos a problemática enfocada, tomamos a psicanálise como nosso referencial teórico, tendo em vista a estreita vinculação entre sujeito e cultura que essa abordagem estabelece. A partir daí, elegemos os seguintes questionamentos como norteadores deste trabalho: considerando a cultura enquanto campo das interdições, como podemos pensar o homem contemporâneo em sua capacidade para desfazer limites ou, simplesmente, negar-lhes a existência? E, a partir da pergunta anterior, como a felicidade é experienciada atualmente, tendo em vista as várias possibilidades de satisfação que nos estão sendo ofertadas? O trajeto que percorremos para discutir tais questões foi: destacamos, inicialmente, a concepção freudiana de homem através das restrições a que este precisa se submeter, tendo em vista sua constituição enquanto um ser culturalizado; em seguida, comentamos como o saber biológico, um dos maiores representantes do discurso científico-tecnológico atual, está se inserindo no âmbito social, reproduzindo a assertiva de que todo limite pode, e deve, ser ultrapassado através das várias possibilidades de se manipular o corpo; em seguida, fizemos a caracterização do homem contemporâneo através de sua busca por satisfações permanentes e que tem o corpo como o alvo preferencial dos prazeres, permitindo, assim, que esse sujeito possa enveredar por vias que lhe asseguram uma posição de excesso; por fim, tendo situado a felicidade em Freud como algo que se refere a escolhas muito próprias ao sujeito em sua relação com o mundo, estabelecemos os caminhos diferenciados seguidos pelo homem contemporâneo em suas altas expectativas no que diz respeito ao ser feliz.

Diss.--Dogmatik--Würzburg--Julius-Maximilians Universität, 2000. / Bibliogr. p. 405-426. Index.

Texte partiel de: Tesi doctoral--Facultat de teologia de Catalunya, 2005. / Bibliogr. p. 59-61.

Thése de doctorat--Vaux-sur-Seine--Faculté libre de théologie évangélique, 2007. / Bibliogr. p. 336-363.