Arquitecturas Flexibles, Crecientes y Jerárquicas para Sistemas Neuronales Autoorganizados

La autoorganización es un proceso de aprendizaje no supervisado mediante el cual se descubren características, relaciones, patrones significativos o prototipos en los datos. Entre los sistemas neuronales autoorganizados más usados destaca el el mapa autoorganizado o SOM (Self-Organizing Map), el cual ha sido aplicado en multitud de campos distintos. Sin embargo, este modelo autoorganizado tiene varias limitaciones relacionadas con su tamaño, topología, falta de representación de relaciones jerárquicas, etc. La red neuronal llamada gas neuronal creciente o GNG (Growing Neural Gas), es un ejemplo de modelo neuronal autoorganizado con mayor flexibilidad que el SOM ya que está basado en un grafo de unidades de proceso en vez de en una topología fija. A pesar de su éxito, se ha prestado poca atención a su extensión jerárquica, a diferencia de muchos otros modelos que tienen varias versiones jerárquicas. El gas neuronal jerárquico creciente o GHNG (Growing Hierarchical Neural Gas) es una extensión jerárquica del GNG en el que se aprende un árbol de grafos, donde el algoritmo original del GNG se ha mejorado distinguiendo entre una fase de crecimiento y una fase de convergencia. Los resultados experimentales demuestran las capacidades de autoorganización y aprendizaje jerárquico de esta red.

Implementación eficiente de redes neuronales autoorganizadas en CUDA C/C++.

En este documento se expondrá una implementación del problema del viajante de comercio usando una implementación personalizada de un mapa auto-organizado basándose en soluciones anteriores y adaptándolas a la arquitectura CUDA, haciendo a la vez una comparativa de la implementación eficiente en CUDA C/C++ con la implementación de las funciones de GPU incluidas en el Parallel Computing Toolbox de Matlab. La solución que se da reduce en casi un cuarto las iteraciones necesarias para llegar a una solución buena del problema mencionado, además de la mejora inminente del uso de las arquitecturas paralelas. En esta solución se estudia la mejora en tiempo que se consigue con el uso específico de la memoria compartida, siendo esta una de las herramientas más potentes para mejorar el rendimiento. En lo referente a los tiempos de ejecución, se llega a concluir que la mejor solución es el lanzamiento de un kernel de CUDA desde Matlab a través de la funcionalidad incluida en el Parallel Computing Toolbox.

Mapping the Liminal Experience of Eastern Europe in Selected American Works, 1965-2002

The purpose of this dissertation is to study literary representations of Eastern Europe in the works of celebrated and less-known American authors, who visited and narrated the region between the mid-1960s and early 2000s. The main critical body focuses on Eastern Europe before 1989 and encompasses three major voices of American literature: John Updike, Joyce Carol Oates, and Philip Roth. However, in the last chapter I also explore American literary perceptions of the area following the collapse of communism. Importantly, the term “Eastern Europe” as used in this dissertation is charged with significance. I approach it not only as a space on the map or the geopolitical construct which emerged in the aftermath of the Second World War, but rather as a conceptual category and a repository of meanings built out of fact and fantasy: specific historical, political and cultural realities interlaced with subjective worldviews, preconceptions, and mental images. The critical framework of this dissertation is twofold. I reach for the concept of liminality to elucidate the indeterminacy and malleability which lies at the heart of the object of study—the idea, image, and experience of Eastern Europe. Bearing in mind the nature of the works under analysis, all of which were inspired by actual visits behind the Iron Curtain, I propose to interpret these transatlantic literary journeys in terms of generative experience, where Eastern Europe is mapped as a liminal space of possibility; a contact zone between cultures and, potentially, the locus of self-discovery and individual transformation. If liminality is the metaphor or a lens that I employ in order to account for the nature of the analyzed works and the complex terrain they map, imagology, whose purpose is to study the processes of constructing selfhood and otherness in literature, provides me with the method and the critical vocabulary for analyzing selected literary representations. The dissertation is divided into six chapters, the last of which serves as coda to the previous discussion. The first two chapters constitute the critical foundation of this work. Then, in chapters 3, 4, and 5 I study American images of Eastern Europe in the works written by John Updike, Joyce Carol Oates, and Philip Roth, respectively. The last, sixth chapter of this dissertation is divided into two parts. In the first one, I discuss new critical perspectives and avenues of research in the study of Eastern Europe following the collapse of communism. Then, I carry out a joint analysis of four works written after 1989 by Eva Hoffman, Arthur Phillips, John Beckman, and Gary Shteyngart. The dissertation ends with conclusions in which I summarize my findings and reflections, and suggest implications for future research. As this dissertation seeks to demonstrate, Eastern Europe portrayed in the analyzed works oscillates between contradictory representations which are contingent upon a number of factors, most importantly who maps it and in what context. Even though each experience of Eastern Europe is distinct and fueled by the profiles, identities, and interests of the characters and their creators, I have found out that certain patterns of othering are present in all the works. Thus, my research seems to suggest that there is something of a recurrent literary image of Eastern Europe, which goes beyond the context of the Cold War. Accordingly, while this dissertation hopes to be a valid contribution to the study of literary and cultural mappings of Eastern Europe, it also generates new questions regarding the current, post-communist representation of the area and its relationship to the national tropes explored in my work.