Paper Details: Downloads: 389
Serial Number: P1120535137
Title: Fuzzy Sarsa: An approach to linear function approximation in reinforcement learning
Authors: L. Tokarchuk and J. Bigham and and L. Cuthbert
Abstract: This paper investigates two different approaches to learning using an agent electronic marketplace as test bed. The types of learning considered in this paper include the temporal difference (TD) learning algorithm Sarsa, and two new fuzzified versions of this algorithm, FQ Sarsa and Fuzzy Sarsa. We implement the three learning algorithms in an agent test bed in order to determine their usefulness in the context of an electronic marketplace. We present the results of various tests demonstrating that the Fuzzy Sarsa algorithm, while having the smallest state space, is also the more effective method of learning.
Keywords: Fuzzy logic, Reinforcement Learning, Agent Systems
Journal/Conference: ICGST Conference on Artificial Intelligence and Machine Learning, AIML-05
Volume:
Issue:
Submission Date: 7/1/2005 12:00:00 AM
Review Date: 10/1/2005 12:00:00 AM
Publishing Date: 12/19/2005 12:00:00 AM
Article Downloads: 389
Download:

Facebook