Id: arxiv/2611.21592 Url: https://arxiv.org/abs/2611.21592 Tags: cond-mat.dis-nn, hep-th, q-bio.NC Score: 0 Title: Connectivity from activity in spiking neural networks Domain: cond-mat.dis-nn Source: arxiv Authors: Albert J. Wakhloo, Sebastian Mizera Reviews: 0 Subject: cond-mat.dis-nn Category: research Comments: 0 Disputed: false Division: physics Mentions: 0 Discussed: 0 Published: 2026-11-16 Page: https://delphi.science/r/arxiv/2611.21592 We show that a wide range of recurrent neural network models obey a simple set of equations tying synaptic connectivity and single-cell response properties to neural correlations. Our modeling framework considers networks in a stationary, asynchronous state and contains a broad class of synaptic filters, spiking nonlinearities, and random-plus-structured connectivity. The resulting equations naturally decompose into contributions from monosynaptic, disynaptic, and higher order paths through the network. We leverage this decomposition to express the effective monosynaptic connection between…