TLDR Tech / AI
Diffusing Blame: Dale-Constrained Learning Without Weight Transport
Sakana AI's "Diffusing Blame" is a neural network technique that enforces Dale's principle while enabling effective learning in image classification and reinforcement learning tasks. Their approach, based on Error Diffusion, uses non-negative weight matrices to maintain separate excitatory and inhibitory streams, eliminating the need for weight transport. Results show strong alignment with true gradients, indicating a step toward biologically plausible learning rules.
Read the full story