← sources

applying Dopamine RL to a static biological wiring diagram (Fruit Fly Connectome) playing Chess.

source: reddit (reddit:reinforcementlearning)published 2026-09-19 19:34ingested 2026-09-20 02:38

retained text

>Instead of using a standard dense neural net, I wanted to see if Reinforcement Learning could carve operational pathways through a rigid, biological matrix.  
>  
>I mapped the Google fruit fly connectome (164,000 neurons) into PyTorch. The electricity can only flow where the biological synapses actually exist. I hit it with virtual dopamine (synaptic weight adjustments) only when it executes a legal chess move or captures a piece.  
>  
>**GitHub:** [https://github.com/AdilSiddiquiHQ/fly-connectome-chess](https://github.com/AdilSiddiquiHQ/fly-connectome-chess)  
>  
>If anyone wants a unique RL environment to train on your local GPU, clone the repo! Click the "Train" button in the web UI. If you get the ELO higher than mine, submit a PR with your `weights.pt`!

document fetched 2026-09-20 00:35 · canonical https://reddit.com/r/reinforcementlearning/comments/1wkvbyh/applying_dopamine_rl_to_a_static_biological