metadata
license: mit
language: en
tags:
- reinforcement-learning
- lunar-lander
- tiny-model
- 394-params
pipeline_tag: reinforcement-learning
LunarLander — 394 params, 95% full landings
A tiny neural network (394 parameters) solving LunarLander-v2. Weights: 1.5 KB.
Performance
| Metric | Value |
|---|---|
| Params | 394 |
| Full landings (reward >= 200) | 95/100 |
| Average reward | 258.8 |
| Median reward | 265.8 |
| Best reward | 310.9 |
| Crashes (reward < 100) | 3/100 |
| Weights size | 1.5 KB |
Demo
Architecture
Params: 8x30 + 30 + 30x4 + 4 = 394