DQN Dribble Ball Simulation (Just for fun, nothing sciency!)
Takes ~30 minutes to train. If you're impatient, scroll down and load the pre-trained net
experience replay size: 2023
exploration epsilon: 1
age: 2025
average Q-learning loss: 0.14619276390858763
smooth-ish reward: 0.798859930566806
Panel
Left: Average reward over time. Right: Network architecture.
Controls
|
I/O
Load:
Download