herurg/Reinforce-Pixelcopter-PLE-v0

REINFORCE Agent playing Pixelcopter-PLE-v0

This is a trained REINFORCE agent for Pixelcopter-PLE-v0, created for Unit 4 of the Hugging Face Deep Reinforcement Learning Course.

Evaluation

• Training episodes: 30000
• Evaluation episodes: 10
• Mean reward: 42.30
• Standard deviation: 20.96
• Course result (mean_reward - std_reward): 21.34
• Required Unit 4 threshold: 5

Hyperparameters

• Hidden layers: 64 → 128
• Gamma: 0.99
• Learning rate: 1e-4
• Max timesteps per episode: 10000
• Algorithm: REINFORCE / Monte Carlo Policy Gradient

The environment was run using a Gymnasium-compatible adaptation of the original Unit 4 notebook.