• It is an extension of Unity ML agent for deep learning, primarily reinforcement learning, with in-editor/in-game training support. It also provides interface for another optimization algorithm called MAES.

• It uses a modified version of KerasSharp and TensorflowSharp as backend. No python is needed for model building/evaluation/training. You can even build a standalone with training capability.

• This repo is made for Aalto University's Computational Intelligence in Games course. The original materials are made with CNTK. Now it is remade with Tensorflow. It will also be my master's thesis hopefully.

• Note: This project is still in development. Don't use this unless you know that you have time to check the sourcecodes!

• Use your already made Unity ML-Agent, but enable learning in Unity editor/build without python or extra coding.
• Reinforcement learning(using PPO) and supervised learning.
• Evolution Strategy (using Matrix Adaption Evolution Strategy(MEAS))
• Examples provided.

For more information including installation and usage instructions, go to Document.

Windows is almost fully supported. If you want to use GPU, CUDA and cuDNN are needed(Please google CUDA v9.0 and cudnn v7 and install them). Mac should be fully supported if I have a Mac to build, but now it does not have Concat Gradient. Mac does not support GPU. Linux is not tested at all.

Android does not support any type of gradient/training. IOS is not tested a all.

We plan to keep this repo updated with latest game related machine learning technologies for the course every year.

Possible future plans/contributions:

• Updating KerasSharp.
• More benchmark environments.
• Better API for in game usage.
• More algorithms including: Deep Q Learning, Deep Mimic, Evolved Policy Gradient, Genetic Algorithms and so on.
• Improving the logging tool or using Tensorboard in c#.
• Graphic editor for neural network architecture

MIT.