Sentient Environment Explorer
A small grid world where an agent has to find food and avoid hazards, and a neural learner works out which moves to repeat.
Independent experiment development. Python, deep Q-learning and simulation UI.
Food, hazards and hunger give movement a consequence, and the environment, the learner and the orchestrator are separate, so a run can be paused and watched while it's still learning.
Saved checkpoints from earlier runs are in the repo. They're not good agents, they're just what came out.
State
A visual reinforcement-learning prototype with controls and checkpoints.