REINFORCE (Monte Carlo Policy Gradient) Agent playing CartPole-v1
This is a trained model of a REINFORCE policy network playing CartPole-v1.
Built and trained in PyTorch with variance reduction baseline normalization.
Model Architecture
- Input: 4 continuous state features (Cart Position, Cart Velocity, Pole Angle, Pole Angular Velocity)
- Hidden Layers: Linear(4, 64) -> ReLU -> Linear(64, 64) -> ReLU -> Linear(64, 2)
- Action Selection: Categorical distribution over discrete actions [Push cart left, Push cart right]
Evaluation Results
- Mean Reward:
9.33 +/- 0.81(Passing threshold: $\ge 350.0$) - Number of Evaluation Episodes: 100
- Model weights: Saved in
reinforce_cartpole.pt
Trained and submitted by Subhash3008.
Evaluation results
- mean_reward on CartPole-v1self-reported9.33 +/- 0.81