Physics > Biological Physics
[Submitted on 25 Jan 2017 (this version), latest version 17 Oct 2017 (v2)]
Title:A complete mean-field theory for dynamics of binary recurrent neural networks
View PDFAbstract:Recurrent networks of binary units provide a prototype for complex interactions in neuronal networks. In the theoretical analysis of their collective dynamics, it is often assumed that unit interactions in the network are homogeneous. Here, we derive a unified theory that encompasses the macroscopic dynamics of recurrent networks with arbitrary connectivity architectures. We identify conditions to the connectivity patterns for which the population averages, according to mean-field theory, converge to a deterministic limit as the size of the system increases. The mathematical analysis, using the martingale structure of the network's Markovian dynamics, further quantifies the aggregate fluctuation of the population activity in a finite-size system. Our theory allows, without additional effort, the investigation of systems that do not exhibit a pointwise convergence to the mean-field limit. In particular, we uncover a novel dynamic state in which a system with inhomogeneous micro-structures in the network connectivity, along with an asynchronous behavior, also exhibits stochastic synchronization. This phenomenon resembles non-trivial fluctuations in the average population activity, which can survive irrespective of system size. The concise mathematical analysis that is presented here enables formulation of mean-field equation for these cases in the form of a stochastic ordinary differential equation. Our results indicate the essential elements for implementation of connectivity motifs in the general theory of recurrent networks.
Submission history
From: Farzad Farkhooi [view email][v1] Wed, 25 Jan 2017 01:23:39 UTC (153 KB)
[v2] Tue, 17 Oct 2017 08:28:45 UTC (145 KB)
Current browse context:
physics.bio-ph
Change to browse by:
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.