A new research paper explores the impact of bias on the absolute capacity of dense associative memory, particularly in the context of Krotov-Hopfield models. The study analyzes how bias in binary patterns affects memory capacity, revealing different asymptotic forms for capacity depending on the order of polynomial interactions and the degree of bias. For fixed bias, the capacity scales as O(N^(n/2)) or O(N^((n+1)/2)) for even or odd interaction orders, respectively, contrasting with the O(N^(n-1)/ln N) capacity observed in unbiased patterns. The research suggests a bias-induced crossover in capacity and proposes an activity-dependent control potential to restore the unbiased capacity. AI
RANK_REASON The cluster contains a research paper detailing theoretical analysis and simulations of associative memory. [lever_c_demoted from research: ic=1 ai=0.7]
- activity-dependent control potential
- arXiv
- bias
- conditioned-Gaussian approximation
- crosstalk mean
- dense associative memory
- Krotov-Hopfield
- patterns
- polynomial interactions
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
