Johann Bernoulli Colloquium

Het Johann Bernouilli Instituut heeft een interessante lezing op het programma staan, over een model voor spiking. Hieronder volgt de abstract:

An important current problem in neuroscience is inferring connectivity in neuronal networks from recordings of spike trains of many neurons. To do this, one chooses a stochastic dynamical model for the network and fits its parameters, including connection strengths, using the data. In the14 approach described here, the model ch14osen to make the fit with is probably the simplest one that captures generic features of the way neurons work: a dynamical Ising (Glauber) model, in which the “up” state represents neuronal firing and the interaction matrix Jij (which need not be symmetric) describes the synapses. Normally in14 theoretical statistical physics we are given a model, specified by some parameters, and the problem is to compute average values of observable quantities. Here we have an inverse problem: We have the measured statistics, and the task is to infer the parameters of the model. This talk will review recent work on this problem, including exact and mean-field algorithms and applying them to simulated and experimental neurobiological data.