Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm

By David Silver et al

Published on Dec. 5, 2017

Read the original document by opening this link in a new tab.

Table of Contents

Abstract
Introduction
Computer Chess History
AlphaZero Algorithm
Training Procedure
Performance Evaluation
Scalability Analysis
Analysis of Chess Knowledge
References

Summary

The paper discusses how the AlphaZero algorithm achieved superhuman performance in chess, shogi, and Go through self-play reinforcement learning. It compares AlphaZero's performance with existing programs like Stockfish and Elmo, showcasing its superiority. The algorithm's scalability and search strategies are analyzed, highlighting its effectiveness. The document also delves into the chess knowledge discovered by AlphaZero, demonstrating its mastery of various chess openings.