A4.3.6 Reinforcement Learning
Reinforcement learning is the third approach to machine learning, and it works quite unlike the others. There is no dataset of labelled examples and no pile of unlabelled data to sort. Instead there is an agent that learns by doing: it acts, sees what happens, and adjusts, much as a child learns to ride a bike or a player learns a new game. This page describes the loop through which an agent and its environment...