Policy Gradient

Experiment with Policy Gradient methods (description), as well as variance reduction.

Current implementation:

• Continuous and discrete environments
• Baseline network for variance reduction

Usage

Setup environment

$ conda env create -f [environment.yml | environment_cuda.yml]
$ conda activate [policy_grad | policy_grad_cuda]

Run training

$ python main.py --config_filename config_filename