'More Is Different' in neural circuits: algebraic emergence of effective theories in recurrent motifs

neurovium · x · 2026-09-04

An arXiv paper by Nima Dehghani models canonical circuit motifs (divisive normalization, winner-take-all) as finite transformation systems and applies Krohn–Rhodes/holonomy decomposition. Key findings: aperiodic primitive updates can compose into non-aperiodic monoids; WTA-to-DN composition yields genuinely composite local cycles certified as group components of the Krohn-Rhodes cascade — an algebraic account of 'More Is Different' in biological neural networks.

Related event: New Paper Uses Krohn-Rhodes Algebra to Explain 'More is Different' in Neural Circuits(2 posts)→

Original post →

More from Research

Research channel →