Saturday, July 27, 2024
HomeRecommendedExplain how to make Shogi AI! Recommended software and books in 2022

Explain how to make Shogi AI! Recommended software and books in 2022

Shogi AI

Isn’t there some people who want to implement AI and make shogi software?

In recent years, Shogi AI has become more active in the shogi world, and shogi battles have complicated patterns, making it difficult to capture against humans. , and has continued to evolve since then.

Some people may want to build their own computer that can play shogi games like this. Therefore, in this article, I will explain how to create an AI that can play shogi.

Table of Contents

  • Can shogi software be made with AI?
  • How to make shogi software
    • Environment
    • Neural network
    • Implementation
  • A book to learn how to make Shogi AI
    • How to create strong shogi software
    • Deep Learning with Shogi AI
    • How to make Shogi AI using deep learning
  • [Recommended for beginners] Shogi software that anyone can make – Shosta
  • Summary

Can shogi software be made with AI?

The rules of shogi are complicated, and it is necessary to predict the future and proceed. Because it is so complicated, many people may find it difficult.

So, is it possible to create shogi software that can play automatically? In conclusion, you can make shogi software that implements AI.

The link below introduces shogi software that uses AI, so if you are interested, please refer to it.

How to make shogi software

Below, I will introduce the flow of actually creating Shogi software. Let’s actually learn game record data and follow the rules so that you can actually play games.

It is mainly divided into the following three processes.

  1. Environment
  2. Neural network
  3. Implementation

I will explain each.

Environment

First, build the environment.

  • Visual Studio Build Tools 2015
  • CUDA
  • cuDNN
  • Chainer
  • Python
  • Cupy
  • Chaine

Let’s install the above environment. Once the environment, such as Visual Studio for typing code and Chainer, which is a Python library, is in place, we collect learning data.

Neural network

Once you’re done learning, it’s time to implement the thinking part of the computer. We need two neural networks to predict shogi moves and evaluate scenes.

The move prediction is performed by the Policy Network, and the scene evaluation is performed by the Value Network. Since it is a deep learning technology originally used for Go games, there are some changes.

Implementation

Once the neural network is implemented, we will implement Shogi AI.

If you use the USI engine, you will be able to play games using the model. Using the neural network part built by using the shogi GUI software “Shogi-dokoro”, it is possible to operate the GUI without creating a GUI.

Also, if you use Shogisho, you can play against humans, humans against engines, engines against engines, and so on.

Let’s create the thinking part of the shogi software and implement it using “Shogi-dokoro”.

A book to learn how to make Shogi AI

If you want to learn more about how to make a shogi AI, we recommend using a book. Here are three books that you can learn how to make.

  1. How to create strong shogi software
  2. Deep Learning with Shogi AI
  3. How to make Shogi AI using deep learning

I will explain each.

How to create strong shogi software

The developer of “GCT Denryu”, which won the 1st and 2nd World Shogi AI Denryu Battle, a computer shogi tournament, explains how to create a stronger shogi AI.

The basic parts of Shogi AI, such as how it works, are explained, and by reading this book, you can actually create Shogi AI with code.

Deep Learning with Shogi AI

From the history of computer shogi and the relationship between deep learning, the overview of computer shogi tournaments and how to participate, the theory that is the premise of the shogi AI to be implemented, and the following three shogi AI that can actually play using deep learning, Python and how to implement it with Chainer is explained.

How to make Shogi AI using deep learning

It is divided into reinforcement learning, large-scale learning/acceleration, and Plicy Network. However, please note that the content is old now and may not work with the latest version of the framework you use.

[Recommended for beginners] Shogi software that anyone can make – Shosta

Making your own shogi AI will be difficult for beginners.

Therefore, I would like to introduce Shogi Software Studio, which is recommended for beginners and allows anyone to create a Shogi AI.

Sho-sta has videos that explain how to use the software in detail, and you can see how to make various things for each purpose.

Summary

In this article, I introduced how to make a shogi AI, a book to learn how to make it, and recommended software.

There are not many machines that know how to make a shogi AI, and there are probably many people who cannot make one even if they want to. However, it becomes possible by using easy-to-understand software and acquiring knowledge from books.

RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Recent Posts

Most Popular

Recent Comments