Reinforcement Learning by Value Gradients

Michael Fairbank

The concept of the value-gradient is introduced and developed in the context of reinforcement learning. It is shown that by learning the value-gradients exploration or stochastic b…