policy network

Terms from Artificial Intelligence: humans at the heart of algorithms

In reinforcement learning including many forms of game play learning, the policy network is a neural network (or ogther form of machine learning algorithm), which attempts to predict the effect of one's actions on the state of the world. In turn taking games, such as Go, this will typically involve predicting the opponent's response to the player's move.

Defined on page 237

Used on page 237