miyosuda/unreal
TensorFlow implementation of DeepMind's UNREAL reinforcement learning algorithm for training agents with auxiliary tasks.
Not currently ranked — collecting fresh signals.
star history
This repository replicates the UNREAL (UNsupervised REinforcement and Auxiliary Learning) algorithm from Google DeepMind’s 2016 paper. It trains agents in the DeepMind Lab 3D environment using deep reinforcement learning with shared convolution and LSTM layers. The implementation includes auxiliary tasks alongside the main reward signal to accelerate learning, demonstrating the approach across various navigation levels with performance graphs tracking agent progress.
Frequently asked
- What is miyosuda/unreal?
- TensorFlow implementation of DeepMind's UNREAL reinforcement learning algorithm for training agents with auxiliary tasks.
- Is unreal open source?
- Yes — miyosuda/unreal is an open-source project tracked on heatdrop.
- What language is unreal written in?
- miyosuda/unreal is primarily written in Python.
- How popular is unreal?
- miyosuda/unreal has 423 stars on GitHub.
- Where can I find unreal?
- miyosuda/unreal is on GitHub at https://github.com/miyosuda/unreal.