Dynamics of Supervised and Reinforcement Learning in the Non-Linear Perceptron

The ability of a brain or a neural network to efficiently learn depends crucially on both the task structure and the learning rule.