Reinforcement Learning Portfolio

• Pong-PPO – Improved PPO with vectorized Atari preprocessing, GAE, and augmentation.

• BipedalWalker – PPO agent with normalized vector envs and deterministic evaluation.

  • 

• Lunar-Landar – Simple policy-gradient style training for LunarLander; latest clip (epoch 526).

training_epoch_526.mp4

• VizDoom-RL – DQN/Double DQN with dueling heads and soft target updates.