Improving Computer Games Performance Using Batch Reinforcement Learning

With the progress of researches in human brain they found that when human learns something new the brain cells structure is changed. That means it is possible to manufacture and develop intelligence. The idea is the same for the device. As intelligence requires knowledge it is necessary for computers to learn and gain knowledge. Machine learning serves this purpose. With the rapid development of computer games we need continuous new techniques because with the increasing numbers of players only a tough games with high policy actions and tactics survive. This book will be useful to researchers who are interested in getting a good idea about computer games. We proposed a new algorithm based on LSPI (Batch Reinforcement Learning Algorithm) called Least-Squares Continuous Action Policy Iteration (LSCAPI). We implemented on two different types of games 8-queens board game and Glest and StarCraft Brood War real-time strategy (RTS) games. The book can have benefit for the beginners in computer games field it gives sequential steps of the game as an idea ending with the implementation.