International Conference on Computational Science, ICCS 2011 Farmer-Pest Problem: A Multidimensional Problem Domain for Comparison of Agent Learning Methods

Bartłomiej Śnieżyński, Jacek Dajda, Marcin Mlostek, Michał Pulchny
2011 Procedia Computer Science  
Learning is often utilized by multi-agent systems which can deal with complex problems by means of their decentralized approach. With a number of learning methods available, a need for their comparison arises. This paper presents initial comparison results for selected algorithms (SARSA, Naïve Bayes, C4.5 and Ripper), which are obtained based on the new multi-dimensional Farmer-Pest problem domain, which is suitable for benchmarking learning algorithms. The results show that supervised learning
more » ... algorithms can be used to generate agent strategy. It appears that for simple environment reinforcement learning algorithm together with Naïve Bayes learning gives best results. Although, in a difficult environment, C4.5 and Ripper are the best.
doi:10.1016/j.procs.2011.04.205 fatcat:mi273mtzzbdndm55vn35422epa