Revisiting Prioritized Experience Replay: A Value Perspective (VER)

This repository contains code from a under-reviewed paper of AAAI 25: Revisiting Prioritized Experience Replay: A Value Perspective. Codes are written in python 3, and provided in jupyter notebook format (.ipynb).

Code for Figure 3 and 4: Results of Q-learning and soft Q-learning in Maze

Q-learning: maze_q_learning.ipynb

soft Q-learning: maze_soft_q_learning.ipynb

Code for Figure 4: Results of DQN and soft DQN in CartPole

DQN: cartpole_dqn.ipynb

soft DQN: cartpole_soft_dqn.ipynb

Code for Figure 6: Valuable Experience Replay (VER) in Atari game

SoftDQN_VER_Seaquest.ipynb