ICML paper says regularized learning often looks Hebbian, while noise turns it anti-Hebbian
burny_tech · x · 2026-07-25
Hebbian learning looks “universal” under regularization
This ICML paper argues that many learning algorithms can be interpreted through a Hebbian lens once regularization is added, and through an anti-Hebbian lens when noise is injected. The authors say this may help explain learning dynamics in biological brains.
- The poster claims regularized learning often aligns with Hebbian updates across different algorithms.
- Noise pushes the alignment toward anti-Hebbian behavior.
- The message is framed as a broader interpretation of how both artificial and biological learning may work.
More from Research
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- UniAction: unified action conditioning lets one world model handle mixed action types — ZeYanjie · 2026-10-27
- Infini-AI Lab open-sources Hermit, a framework for agent harness evolution and continual learning — BeidiChen · 2026-10-27
- Meta FAIR's Self-Direction turns agent failures into reusable LoRA skills, +13.8pp on Terminal-Bench — karen_ullrich · 2026-10-27
- DONIC hits 86.79% on bimanual dexterous tracking, crushing prior SOTA's 9.43% — liuziwei7 · 2026-10-27
- Why don't more LLMs ship 10M-token context? A researcher asks about attention tradeoffs — agarwl_ · 2026-10-27