Shogi software has been developed using deep learning, and I think many people are interested in it. AI has beaten a professional shogi player in a match, so I’m very interested in shogi software.
What is the difference from conventional software? Is there software that can be used for free? I hope that through this article, you will know.
Table of Contents
- What is shogi software that uses deep learning?
- What is the difference between deep learning shogi software and conventional shogi software?
- Features of conventional shogi software
- Features of Deep Learning Shogi Software
- Professional shogi players study with deep learning shogi software
- Akira Watanabe introduces deep learning shogi software
- Utilizing Sota Fujii Quadruple Crown, dlshogi and Mizusho
- Is there a free version? Two major deep learning shogi software currently available
- dlshogi-deep learning software
- Mizusho – The Strongest Shogi Software in 2021
- What you need to create your own shogi software with python
- “Sho Star -Shogi Software Studio-“
- The book “How to create a strong shogi software” is recommended
- in conclusion
What is shogi software that uses deep learning?
There are mainly two types.
- Software for playing
- Software to record
In the game software, you can set the level and play against each other, and since it has an analysis function, you can get advice.
Recording software is used to record more than 3,000 games per year. It was introduced in the form of supplementing the personnel responsible for taking notes. A camera is installed on the ceiling, and all pieces are recognized and recorded by image recognition technology that utilizes deep learning.
The following article explains the AI used in shogi. If you are interested, please refer to it.
What is the difference between deep learning shogi software and conventional shogi software?
There have been many shogi software before. But what is the difference between conventional shogi software and deep learning shogi software?
- Features of conventional shogi software
- Features of Deep Learning Shogi Software
We will focus on each feature and explain.
Features of conventional shogi software
As a feature of AI, it is not influenced by human emotions. Humans can not predict the future, and in the case of a high-risk hand, it is greatly influenced by the psychological state.
A feature of AI is that it can point the battle situation in an advantageous way with a statistical move based on data without being influenced by the state of mind.
Conventional shogi software looks ahead and searches for as many positions as possible. It uses human-understandable feature values and uses machine learning to give points.
Features of Deep Learning Shogi Software
Unlike previous shogi software, it is not possible to check the position very much. The number of opposing phases that can be read in one second is small.
However, deep learning shogi software has a very high accuracy in position evaluation, so it has the advantage of being strong even without reading the number of positions.
Professional shogi players study with deep learning shogi software
Akira Watanabe introduces deep learning shogi software
Shogi master Akira Watanabe has purchased a computer that can read 80 million hands per second, introduced deep learning software, and is conducting research.
In 2007, Akira Watanabe, who fought computer shogi Bonanza, decided to purchase software using deep learning in order to find the best research method for creating a stronger shogi AI.
Utilizing Sota Fujii Quadruple Crown, dlshogi and Mizusho
Sota Fujii has made it public that he uses Mizusho when practicing Shogi.
However, it seems that Sota Fujii started using dlshogi right after dishogi won the Denryu game (computer shogi tournament).
Both apps are pretty strong, so using both is the right idea.
Is there a free version? Two major deep learning shogi software currently available
There are many shogi software for PC.
- Dlshogi
- Water master
Among them, it can be said that “Mizusho” and “dlshogi” are generally the top two. We will explain each below.
dlshogi-deep learning software
dlshogi is a deep learning-based shogi AI developed by Tadao Yamaoka. Due to the large number of file settings, it is said that the difficulty of introduction is high. It may be difficult for people who are not familiar with computers.
The Windows version file of dlshogi (2nd World Shogi AI Denryu Battle Exhibition Version) is open to the public and can be downloaded from the link below.
Be sure to check the license of the model file before downloading it.
Mizusho – The Strongest Shogi Software in 2021
Mizusho is software based on King Yaneura.
This is the so-called “NNUE system”, a conventional type of software that runs on a CPU, and can demonstrate high performance even on a general PC. Mizusho can also be downloaded for free, but it requires the use of a GUI tool called Shogi-dokoro.
The Mizusho file is included in the Yaneura Oh file, so please download it from the link below.
What you need to create your own shogi software with python
I would like to introduce two main aspects of development.
- Sho Star -Shogi Software Studio
- Book “How to Create Strong Shogi Software”
I will explain each one.
“Sho Star -Shogi Software Studio-“
“Shostar – Shogi Software Studio” is a tool that can create learning data for Shogi AI and Shogi software. Sueyoshi. It is published by Mr. and can be downloaded for free.
Learning data can be generated based on the score and rating of “floodgate”, in which shogi programs play consecutive games. In addition, you can create Shogi software simply by specifying the software name, author name, learning model, etc.
Both are characterized by being easy to operate without programming knowledge. You can download it from the link below. There is also a video on how to use it, so please refer to it.
The book “How to create a strong shogi software” is recommended
In “How to create strong shogi software”, he is good at creating shogi applications that implement deep learning, and has a comprehensive explanation.
It consists of four parts and is explained in an easy-to-understand manner.
Chapters 1 and 2 explain the algorithms used in computer shogi. By using deep learning, we explain where it is superior to conventional software.
In chapters 3 and 4, we will implement it using Python. In the development environment, we use Google Colab notebooks, which can be used for free, so that we can create them smoothly. It also explains how to make stronger software.
In conclusion
I have explained about shogi software that implements deep learning. It is very surprising that even professional shogi players use such software to study shogi.
As deep learning continues to be used in the future, amazing software may be developed. If you want to make your own shogi software, there are books on how to actually make software, so why not use them as a reference?
