Análisis comparativo de la capacidad de aprendizaje del paradigma CBR respecto a otros modelos de aprendizaje máquina
Loading...
Download
Official URL
Full text at PDC
Publication date
2024
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
En este trabajo se abordará el análisis de distintas técnicas del campo del aprendizaje de máquina, centrándonos específicamente en el paradigma de Caso Basado en la Razón (CBR) y comparándolo con otros modelos de aprendizaje, como el aprendizaje por refuerzo. El objetivo principal es realizar un análisis exhaustivo de la capacidad de aprendizaje de CBR en comparación con otros enfoques, explorando sus fortalezas y limitaciones. Para ello hemos utilizado el juego MsPacMan Vs. Ghost como entorno de evaluación de distintos comportamientos de aprendizaje automático. En él se han aplicado y evaluado diversas implementaciones utilizando diversas técnicas de aprendizaje automático. Las técnicas utilizadas han sido: algorítmica, máquinas de estados finitos, sistemas basados en reglas, sistemas con razonamiento basado en casos y aprendizaje por refuerzo. Para cada técnica se ha analizado cómo están implementadas y su funcionamiento. Posteriormente se han realizado una serie de evaluaciones entre distintas técnicas y se han analizado los resultados. Concretamente se han analizado los resultados de evaluar las implementaciones CBR y Q-learning contra el resto de las técnicas elegidas. A lo largo del desarrollo del proyecto se han ido sacando conclusiones y se han visto las ventajas e inconvenientes de cada técnica. El código de este proyecto es accesible a través de este enlace: https://github.com/Dbugoi/TFM
This paper will address the analysis of different techniques in the field of machine learning, focusing specifically on the Case Based Reasoning (CBR) paradigm and comparing it with other learning models, such as reinforcement learning. The main objective is to perform a comprehensive analysis of the learning capability of CBR in comparison with other approaches, exploring its strengths and limitations. For this purpose, we have used the MsPacMan Vs. Ghost game as an environment to evaluate different machine learning behaviors. Several implementations have been applied and evaluated using different machine learning techniques. The techniques used were: algorithmic, finite state machines, rule-based systems, case-based reasoning systems and reinforcement learning. For each technique we have analyzed how they are implemented and how they work. Subsequently, a series of evaluations have been carried out between different techniques and the results have been analyzed. Specifically, the results of evaluating the CBR and Q-learning implementations against the rest of the chosen techniques have been analyzed. Throughout the development of the project, conclusions have been drawn and the advantages and disadvantages of each technique have been seen. The code for this project is accessible through this link: https://github.com/Dbugoi/TFM
This paper will address the analysis of different techniques in the field of machine learning, focusing specifically on the Case Based Reasoning (CBR) paradigm and comparing it with other learning models, such as reinforcement learning. The main objective is to perform a comprehensive analysis of the learning capability of CBR in comparison with other approaches, exploring its strengths and limitations. For this purpose, we have used the MsPacMan Vs. Ghost game as an environment to evaluate different machine learning behaviors. Several implementations have been applied and evaluated using different machine learning techniques. The techniques used were: algorithmic, finite state machines, rule-based systems, case-based reasoning systems and reinforcement learning. For each technique we have analyzed how they are implemented and how they work. Subsequently, a series of evaluations have been carried out between different techniques and the results have been analyzed. Specifically, the results of evaluating the CBR and Q-learning implementations against the rest of the chosen techniques have been analyzed. Throughout the development of the project, conclusions have been drawn and the advantages and disadvantages of each technique have been seen. The code for this project is accessible through this link: https://github.com/Dbugoi/TFM
Description
Trabajo de Fin de Máster en Ingeniería Informática, Facultad de Informática UCM, Departamento de Ingeniería de Software e Inteligencia Artificial, Curso 2023/2024.
El código de este proyecto es accesible a través de este enlace: https://github.com/Dbugoi/TFM