Discount Factor as a Regularizer in Reinforcement Learning

Code for the paper "Discount Factor as a Regularizer in Reinforcement Learning", Ron Amit, Ron Meir, Kamil Ciosek, ICML 2020. [Paper] [Slides] [Video] [bib]

DDPG and TD3 code is based on: https://github.com/sfujim/TD3

Basic Requiremnts

• Python 3.7
• NumPy, Matplotlib, and seaborn
• Ray 0.84

Requiremnts for Mujoco Experiments

• PyTorch 1.5.1 with CUDA
• Gym 0.17.1
• Mujoco 1.50
• mujoco-py 2.0.2.8

Reproducing Results

All results are saved in this zip file (including complete parameters, raw results, and figures).