Curiosity-Driven Reinforcement Learning Through Temporal Distance

Foundations of Machine Learning 2018 Project by Dae Young Kim

PPO implementation code in PyTorch imported from PyTorch PPO.

My modifications/writings are only in:

• README.md
• main.py
• mymodels.py
• plottheresults.py
• ./a2c_ppo_acktr/arguments.py
• ./plots/rewardbp.py

Requirements:

• PyTorch
• TensorFlow
• Open AI Gym
• pybullet

To run:

python main.py --env-name 'BipedalWalkerHardcore-v2' --use_tdm True --beta_int 10.0 --num_layers 2 --fc_width 300 --opt_lr 1e-4 --beta_schedule linear --bonus_func log

Plots: