Stochastic Synaptic Dynamics Under Learning
Open Access
- 13 April 2026
- journal article
- Published by American Physical Society (APS) in PRX Life
- Vol. 4 (2)
Abstract
Learning is based on synaptic plasticity, which affects and is driven by neural activity. Because pre- and postsynaptic spiking activity is shaped by randomness, the synaptic weights follow a stochastic process, requiring a probabilistic framework to capture the noisy synaptic dynamics. We consider a paradigmatic supervised learning example: a presynaptic neural population impinging in a sequence of episodes on a recurrent network of integrate-and-fire neurons through synapses undergoing spike-timing-dependent plasticity (STDP) with additive potentiation and multiplicative depression. We first analytically compute the drift and diffusion coefficients for a single synapse within a single episode (microscopic dynamics), mapping the true jump process to a Langevin and the associated Fokker-Planck equations. Leveraging recent analytical tools, we include spike-time-resolving cross-correlations between pre- and postsynaptic spikes, which corrects substantial deviations seen in standard theories purely based on firing rates. We then apply this microdynamical description to the network setup in which hetero-associations are trained over one-shot episodes into a feed-forward matrix of STDP synapses connecting to neurons of the recurrent network (macroscopic dynamics). By mapping statistically distinct synaptic populations to instances of the single-synapse process above, we self-consistently determine the joint neural and synaptic dynamics and, ultimately, the time course of memory degradation and the memory capacity. We demonstrate that specifically in the relevant case of sparse coding, our theory can quantitatively capture memory capacities which are strongly overestimated if spike-time–resolving cross-correlations are ignored. On the single-synapse level, our theory predicts a transition between acceleration and deceleration of the synaptic weight, and a possibly nonmonotonic dependence of the finite-time diffusion coefficient on the synaptic weight. On the memory level, we showcase two complementary roles of homeostatic plasticity: namely, it is both responsible for forgetting and a requirement for the ability to learn. Additionally, we show how strength of plasticity and duration of exposure to stimuli must be scaled to evade a tradeoff dilemma between memory capacity and early accuracy. We conclude with a discussion of the many directions in which our framework can be extended.Keywords
All Related Versions
Funding Information
- Deutsche Forschungsgemeinschaft (327654276)
This publication has 67 references indexed in Scilit:
- Analysis and Data-Based Reconstruction of Complex Nonlinear Dynamical Systems2019
- Exact analytical results for integrate-and-fire neurons driven by excitatory shot noiseJournal of Computational Neuroscience, 2017
- Neural associative memories and sparse codingNeural Networks, 2013
- Neural Mechanisms of Speed-Accuracy TradeoffNeuron, 2012
- Coordination of size and number of excitatory and inhibitory synapses results in a balanced structural plasticity along mature hippocampal CA1 dendrites during LTPHippocampus, 2011
- Emergence of network structure due to spike-timing-dependent plasticity in recurrent neuronal networks. I. Input selectivity–strengthening correlated input pathwaysBiological Cybernetics, 2009
- Spike-train spectra and network response functions for non-linear integrate-and-fire neuronsBiological Cybernetics, 2008
- Are the input parameters of white noise driven integrate and fire neurons uniquely determined by rate and CV?Journal of Theoretical Biology, 2008
- Phenomenological models of synaptic plasticity based on spike timingBiological Cybernetics, 2008
- Handbook of Stochastic Methods for Physics, Chemistry and the Natural Sciences1985