REINFORCE Agent — Pixelcopter-PLE-v0
This model was trained using the REINFORCE Monte Carlo Policy Gradient algorithm.
Environment
Pixelcopter-PLE-v0
Training
Training episodes: 50,000
Hidden size: 64
Learning rate: 0.0001
Gamma: 0.99
Evaluation
Mean reward: 58.20
Standard deviation: 46.93
Certification result:
11.27
The certification result is calculated as:
mean reward - standard deviation
Course
Hugging Face Deep Reinforcement Learning Course
Unit 4 — Policy Gradient Methods
Evaluation results
- mean_reward on Pixelcopter-PLE-v0self-reported58.20 +/- 46.93