New learning rule boosts Hopfield network memory capacity over 10×, per Nature Communications study
bravo_abad · x · 2026-11-12
- Blümel et al. (Nature Communications, 2026) show that changing only the learning rule gives a classic Hopfield network over 10× more storage capacity.
- A Hopfield network stores patterns and recalls them from a cue, but pattern interference limits capacity. The new objective maximizes the shared information between the two signals each neuron receives during training—the pattern to store and recurrent input from other neurons.
- On random-pattern benchmarks this yields 1.59 stored patterns per neuron vs 0.14 for the classical Hebb rule (memory capacity limit shifts from αH to αC in the figure).
- The objective was derived from what successful storage already had in common: high information redundancy. The transferable move: turn a measurable explanation of why a model works into a training objective.
More from Research
- PhyRubric: rubric-based benchmark diagnoses physical commonsense in text-to-video — Huiyi Chen · 2026-11-12
- Economist: 3 reasons letting AI write your paper abstract is a mistake — Afinetheorem · 2026-11-12
- CMU/JetBrains paper: engineer token spend measures AI usage, not output shipped — rohanpaul_ai · 2026-11-12
- CMU and JetBrains study: doubling AI token spend yields only ~10% more commits — rohanpaul_ai · 2026-11-12
- ADJ: modeling LLM jailbreaking as a Hegelian-dialectic Stackelberg game between two LLMs — chaumian · 2026-11-12
- AI-Designed T-Cell Receptors Boost Cancer Killing in Lab and Tumor Control in Mice — WmHaseltine · 2026-11-12