Estimation of neural models from spikes

Abstract

In this work we explored traditional methods like particle filters and modern simulation-based inference ones to estimate the parameters and the hidden variables of neuronal models from neuronal spike train responses. From our experiments we can say that modern simulation-based inference methods are able to show similar results than more traditional particle filters methods but within a fraction of the computational time. Moreover, simulation-based inference methods can easily be applied to more complex models and still remain fast and fairly accurate.