If you have been fascinated by the DeepMind AlphaGo program, there is good news for you. Some Go enthusiasts have replicated the results of the Papel AlphaGo Zero, using some resources provided by Google.
Source: WIRED
The developers want to emphasize that this project is in no way associated with DeepMind's official AlphaGo program. It is an independent effort inspired by AlphaGo, but not affiliated with it.
According to the developers, Minigo “is a pure Python implementation of an AI Go based on neural networks, using TensorFlow ". Adds some features and architecture changes to the “Mastering the game of Go without human knowledge” paper. Very recently, this architecture was further expanded in the “Master chess and shogi through self-play with a general reinforcement learning algorithm” paper.
The objectives of this project, described by the authors, are listed below:
- Provide a clear set of learning examples with Tensorflow, Kubernetes and Google Cloud Platform to establish reinforcement learning channels on various hardware accelerators.
- Reproduce the methods of the original DeepMind AlphaGo articles as faithfully as possible, through an open source implementation and open source pipeline tools.
- Provide our data, outdoor findings and discoveries to benefit Go communities, machine learning and Kubernetes.
If you are interested in doing this on your machine, you can access the Minigo source code and other resources, here.
Just a note here that you will need the following before you can get started:
Our opinion on this
The developers repeatedly mention that this is not a DeepMind project and is explicitly not intended to compete with AlphaGo. They just wanted other developers in the community to understand (and maybe they will replicate or improve) how the Go model works. It's definitely something to keep an eye out for as this study progresses further..
