NAME

Generalizable Q-Learning Framework - Deep Q-learning framework that trains agents for Snake, Tic-Tac-Toe, Connect-N, and CartPole

SYNOPSIS

2018; Python, Keras, TensorFlow, OpenAI Gym, Matplotlib
github

DESCRIPTION

Python framework built around a reusable Double DQN agent with a target network, replay memory, Huber loss, and epsilon decay, plus a shared game interface so new environments plug in as small drivers. Supports self-play training for two-player games and ships trained Tic-Tac-Toe weights and Snake reward curves.

HIGHLIGHTS

• Reusable Double DQN agent with target network, replay memory, and configurable hyperparameters shared across games
• Self-play training loop for adversarial games such as Tic-Tac-Toe and Connect-N
• Training diagnostics: moving-average reward plots, outlier removal, and SIGINT-safe stats saving
• Distance-based reward shaping for Snake, with reward and score curves included

SEE ALSO

* projects(7), all projects
* ari(7), the resume
* PDF Resume: Download Resume