Nat CommunJClub
Redundancy maximization as a principle of associative memory learning in Hopfield networks.
Nature Communications · · Journal Article · Open access
Blümel, Schneider + more
Abstract ↗AI summary
The abstract is read at the publisher; the summary is JClub's.
Below Hopfield memory capacity, a neuron’s activity is dominated by high redundancy between external pattern input and internal recurrent input, while synergy and unique information stay near zero until capacity is exceeded.
- Why it matters: The local information-processing rules enabling associative recall in Hopfield networks are not well characterized, limiting principled design of higher-capacity memory models.
- What they did: Using Partial Information Decomposition on individual neurons, the study quantifies unique, redundant, and synergistic information from external and recurrent inputs, then trains networks with redundancy maximization per neuron.
- The result: Redundancy maximization raises memory capacity to 1.59 versus 0.14 for classical Hopfield networks, and it also outperforms recent state-of-the-art Hopfield implementations.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.
- Open access