Rhythmogenesis and functional stability in recurrent networks via STDP

by Mr. Gabi Socolovsky

Department of Physics, Ben-Gurion University of The Negev

at Biological and soft-matter physics

Thu, 23 Jan 2025, 12:10
***ZOOM only***

Abstract

Rhythmic activity has been widely observed in the central nervous systems of a diverse range of animals and was linked to various cognitive tasks. Numerous computational studies investigated models capable of generating such rhythmic activity. In these models, for rhythmic activity to emerge, the strengths of the synaptic connections– synaptic weights, must be within a certain range of values. This raises a fundamental question: How are synaptic weights tuned to maintain rhythmic activity? Furthermore, empirical evidence indicates that synaptic strengths exhibit significant volatility, fluctuating by up to 50% within a few days. This observation introduces another critical challenge: How can rhythmic functionality be preserved despite ongoing synaptic motility?

In this talk, I will present our investigation into the hypothesis that rhythmic activity in neural circuits can emerge and stabilize through activity-dependent plasticity, specifically in the form of spike-timing-dependent plasticity (STDP). We analyze the dynamics of STDP in large recurrent networks in two stages. First, we derive the effective dynamics governing the order parameters of synaptic connectivity. Subsequently, we employ a perturbative approach to assess the stability of these dynamics.

We find that for a wide range of parameters STDP can induce rhythmogenesis. Moreover, STDP can suppress synaptic fluctuations that disrupt functionality. Interestingly, we show how STDP can channel fluctuations in the synaptic weights into a manifold on which the network activity is not affected; thus, maintaining functionality while allowing a subspace in which synaptic weights can be widely distributed.

Zoom Meeting
https://us02web.zoom.us/j/88453495692?pwd=cYb0y6JTYecbqpPewdb4H47LWoajHF.1

Meeting ID: 884 5349 5692
Passcode: 345123Join

Note: graduate student thesis seminar

Created on 01-01-2025 by Granek, Rony (rgranek)
Updaded on 22-01-2025 by Granek, Rony (rgranek)