Google AI Open Sources TensorFlow Code & Dataset for Reinforcement Learning
- Google AI’s team has open sourced a TensorFlow-based framework to reinforcement learning
- The framework is called Dopamine, and the entire code is available for download on GitHub
- Along with this, the team has released the entire training data to help you benchmark your test results
Progress in reinforcement learning has chugged along at a far slower rate than deep learning. While there have been notable news-worthy breakthroughs like OpenAI Five and Google’s AlphaGo, getting reinforcement learning practices into practical scenarios has just not happened.
As Google AI’s team mentions in this blog post, developing these kind of algorithms requires tons of experimentation without any clear direction. Unfortunately most of the existing frameworks out there just do not have that kind of flexibility. If you’ve worked or researched in this field, you know exactly how difficult (if not impossible) it is to reproduce existing approaches.
So to help speed along research and with the hope of getting the community more involved in reinforcement learning, the Google AI team has open sourced a TensorFlow framework, called Dopamine, that aims to create research by making it more flexible and reproducible. According to the team’s official documentation, their design principles are:
- Easy experimentation:To help new users run benchmark experiments
- Flexible development: To facilitate the generation of new and innovative ideas for new users
- Compact and reliable: Provide implementations for some of the older and more popular algorithms
- Reproducible: Ensure results are reproducible
Realizing how important is for new folks to check their results against a benchmark, the researchers have also released the entire training data. It is available as Python pickle files, JSON files, and a website where users can visualize each training iteration.
The code available on GitHub is just 15 Python files and comes packaged with detailed documentations. What are you waiting for? Get started on this NOW!
Our take on this
Note that DeepMind’s research on dopamine was unrelated to this work by Google AI. While both are rooted in reinforcement learning to quite an extent, Google AI has involved the entire community by open sourcing it’s efforts. It certainly helps that it’s TensorFlow based, a framework everyone in the deep learning community is familiar with.
Reinforcement learning can be a daunting subject to start with but I encourage you all to give this a try. This is a field that is still ripe with potential and will see a lot of progress in the coming years. This is a great resource to get started and you can also refer to to our article for beginners.
Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!