Aplicación de técnicas de optimización estadística enfocada a la computación gráfica : modelo de procesamiento de Imágenes conectado a un Agente Q-Learning.

Abstract

This project explores the development and analysis of a Q-Learning agent using convolutional neural networks for graphical environments, specifically Atari video games. The primary objective was to demonstrate how advanced optimization and deep learning techniques can enhance decision-making and interaction within these environments. A comprehensive study of various training strategies was conducted, and the results were analyzed in terms of the agent’s performance and robustness. The results show that employing advanced optimization techniques and convolutional neural networks enables the agent to make more effective decisions in complex environments such as Atari games. This work emphasizes the importance of integrating interactive resources, such as video games, into STEM education, providing a more dynamic and effective learning experience. Additionally, it is highlighted that while complex graphical environments add appeal to the project, the Q-Learning methods employed represent cutting-edge technology with significant applications in artificial intelligence