Firing rate distributions in plastic networks of spiking neurons
In recurrent networks of leaky integrate-and-fire (LIF) neurons, mean-field theory has proven successful in describing various statistical properties of neuronal activity at equilibrium, such as firing rate distributions. Mean-field theory has been applied to networks in which either the synaptic weights are homogeneous across synapses and the number of incoming connections of individual neurons is heterogeneous, or vice versa. Here we extend the previous mean-field formalisms to treat networks in which these two sources of structural heterogeneity occur simultaneously, including networks whose synapses are subject to plastic, activity-dependent modulation. The plasticity in our model is mediated by the introduction of one spike trace per neuron: a chemical signal that is released every time the neuron emits a spike and which is degraded over time. The temporal evolution of the trace is controlled by its degradation rate $r_p$ and by the neuron's underlying firing rate $\nu$. When the ratio $\alpha=\nu / r_p$ tends to infinity, the trace can be rescaled to be a reliable estimation of the neuron's firing rate. In this regime, the value of any synaptic weight at equilibrium is a function of the pre- and post-synaptic firing rates, and this relation can be used in the mean-field formalism. The solution to the mean-field equations specifies the firing rate and synaptic weight distributions at equilibrium. These equations are exact in the limit of reliable traces but they already provide accurate results when the degradation rate lies within a reasonable range, as we show by comparison with simulations of the full neuronal dynamics in networks composed of excitatory and inhibitory LIF neurons. Overall, this work offers a way to explore and better understand the way in which plasticity shapes both activity and structure in neuronal networks.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Normalisation of Weights and Firing Rates in Spiking Neural Networks with Spike-Timing-Dependent Plasticity
Maintaining the ability to fire sparsely is crucial for information encoding in neural networks. Additionally, spiking homeostasis is vital for spiking neural networks with changing numbers of weights and neurons. We dis…
ClusteringTime SeriesTime Series AnalysisReinforcement learning with a network of spiking agents
Neuroscientific theory suggests that dopaminergic neurons broadcast global reward prediction errors to large areas of the brain influencing the synaptic plasticity of the neurons in those regions. We build on this theory…
reinforcement-learningReinforcement LearningReinforcement Learning (RL)Natural-gradient learning for spiking neurons
In many normative theories of synaptic plasticity, weight updates implicitly depend on the chosen parametrization of the weights. This problem relates, for example, to neuronal morphology: synapses which are functionally…
Beyond Pairwise Plasticity: Group-Level Spike Synchrony Facilitates Efficient Learning in Spiking Neural Networks
Brain networks rely on precise spike timing and coordinated activity to support robust and energy-efficient learning. Inspired by these principles, spiking neural networks (SNNs) are widely regarded as promising candidat…
Learning in Spiking Neural Networks with a Calcium-based Hebbian Rule for Spike-timing-dependent Plasticity
Understanding how biological neural networks are shaped via local plasticity mechanisms can lead to energy-efficient and self-adaptive information processing systems, which promises to mitigate some of the current roadbl…
Edge-computing