Using multi-agent systems for machine learning

The interaction between computer systems and the learning that some of them are able to achieve is vital, the changes are visible in the traditional way of analyzing and developing them. The need for interaction between system components is increasingly important in resolving joint tasks that would be individually very costly or even impossible. Multiagent systems offer a broad platform for performing distributed tasks that cooperate with each other, but also allow for the inclusion in each agent of intelligent behavior that can be developed through automatic learning techniques. The reinforcement learning method inserted in automatic learning is very useful for use with agents, allows agents to learn through the interaction of samples and errors in a dynamic environment. This document addresses concepts, characteristics, relationships and other important aspects of the use of multiagent systems for automatic learning.