Backprop-Free Learning Adheres to Dale's Principle
chris_j_paxton · x · 2026-07-19
Quoting a new paper, Diffusing Blame, this post introduces a routing mechanism in deep learning that broadcasts error signals directly to hidden layers. This trains networks that comply with Dale 原则 (neurons either excite or inhibit their neighbors, but cannot do both), thereby eliminating traditional backprop.\n\nThe author notes that this method performs surprisingly well on image recognition and achieves competitive, backprop-free reinforcement learning results in complex motor control tasks and the open-ended Craftax environment.
More from Research
- Project APE finds verifier reliability drops when papers contain multiple errors — soumitrashukla9 · 2026-07-22
- Project APE says verifier costs fell about 90x in a year as Chinese open models lead — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Project APE builds its verifier benchmark from 100 AI-written papers with injected errors — soumitrashukla9 · 2026-07-22
- Paper proposes a CRED taxonomy and benchmark to measure research-error detectors — soumitrashukla9 · 2026-07-22
- OpenAI says long-horizon models need safety and alignment checks across full action sequences — rhiever · 2026-07-22