New theory analyzes stability of memory patterns in neural networks

Researchers have developed a new theory to analyze the dynamical stability of stored patterns in attractor neural networks, which are models of biological memory. This theory extends previous approaches by considering graded neural activities and the presence of noise, using methods from random matrix theory. The study identifies a "critical load for stability" that determines whether stored patterns are stable, a concept distinct from the classical critical capacity. The findings suggest that sparse-like patterns and threshold-linear activation functions offer computational benefits and provide testable predictions for neural circuits. AI

RANK_REASON Academic paper published on arXiv detailing a new theory for neural networks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• condensed matter physics
• DagsHub
• Disordered Systems and Neural Networks
• Gotit.pub
• Uri Cohen

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →