Stability by gating plasticity in recurrent neural networks

Abstract

With Hebbian learning who fires together wires together, well-known problems arise. Hebbian plasticity can cause unstable network dynamics and overwrite stored memories. Unstable dynamics can partly be addressed with homeostatic plasticity mechanisms. Unfortunately, the time constants of homeostatic mechanisms required in network models are much shorter than those measured experimentally. We propose that homeostatic time constants can be slow if plasticity is gated. We investigate how gating plasticity influences network stability and memories in plastic balanced spiking networks of neurons with dendrites. We compare how different factors such as excitability, learning rate, and inhibition lift the requirements for homeostatic time constants. We investigate how dendritic versus perisomatic gating allows for different amounts of weight changes in stable networks. We suggest that the compartmentalisation of pyramidal cells enables dendritic synaptic changes while maintaining stability. We show that spatially restricted plasticity improves stability. Finally, we compare how different gates protect memories. Significance statementHow does the brain maintain stable neural activity in the presence of synaptic changes? This question has been studied extensively in the past, but we argue that one crucial aspect is missing in previous studies. While all theoretical work has assumed plasticity to be on all the time, plasticity is in fact heavily gated. In this light, we must reconsider the theories on stability and homeostasis of neural activity. In particular, theoretical studies show that neural networks undergoing plasticity require fast compensatory homeostatic mechanisms to be stable. However, experimentally measured homeostatic processes operate on much slower time scales. We studied how the gating of plasticity can improve network stability and thereby reduce the discrepancy in the homeostatic time constant between models and experiments.