Learning to Play Tic-Tac-Toe with Jax: Training a DQN via Reinforcement Learning
TL;DR A tutorial shows how to build and evaluate a simple Deep Q Network (DQN) that learns Tic-Tac-Toe using Jax…
Scaffolding to Superhuman: Curriculum Learning Beats 2048 and Tetris
TL;DR An individual researcher used PufferLib, extensive hyperparameter sweeps and curated curricula to train agents that outperform a terabyte-scale search…