Implementation of ACKTR algorithm.
The code is organized as follows:
- models.py: implementations of A2C and A3C
- kfac.py: implementation of K-FAC optimizer
- ac_nets.py: neural network architectures of actor and critic for different environments
- storage.py: implementation of data structure to efficiently store information during learning
- utils.py: utils for models and optimizer
See demo here.
Code is developed and supported by:
- Iurii Kemaev hbq1 (y.kemaev@gmail.com)
- Maxim Kuznetsov binom16 (binom16@gmail.com)
- Eugenii Nikishin nikishin-evg (nikishin.evg@gmail.com)