April 2018
Intermediate to advanced
334 pages
10h 18m
English
In Chapter 5, Q-Learning and Deep Q Networks, we studied game trees and minimax algorithms. Let's recall those approaches to understand the architecture behind the AI program of IBM DeepBlue.
A game tree represents the full end-to-end representation of a game, where each node in the tree represents a particular game state (position) and the edges linking the nodes represent the moves (actions) taken at previous game states leading to a new game state. The root node represents the start of the game and nodes in the next level represent the possible states generated after all different possible actions have been taken at the start state of the game and similarly nodes in the further layers are generated. ...