Reinforcement Learning

Reinforcement Learning

Reinforcement learning essentially boils down to trying actions, recording the results of those actions, and repeating actions that turned out favorable. It generally fits into the 'unsupervised learning' category, but this strategy still requires that 'favorability' be pre-defined. This definition can come directly from a human, from an evolutionary algorithm, etc. but as far as the reinforcement learning process is concerned, 'favorability' is a given.

In the early stages of spider's intelligence, this is defined in code, but in later stages it may be defined by some sort of 'genome' that defines a number of features for each individual spider.