|
Abstract:
|
How neural population preserves multiple simultaneously presented stimuli is a fundamental question, but has not yet been fully understood by neuroscientists. Related studies suggest a single neuron may present a trial-wise multiplexing and a time division `Mixture' dynamic under dual stimuli, where a single neuron may encode information by fluctuating between two stimuli from trials to trials. Here we are interested in whether or how such turn-taking activities may present a coordination pattern in neural population. We define a fluctuate weight to measure selection preference for a single neuron at a trial by a Poisson mixture model. To capture the neural population coordination, we introduce a block correlation matrix of fluctuating weights by a Gaussian copula model. We estimate the block correlation matrix with unknown cluster assignment in a Bayesian framework. Specifically, we consider a canonical representation of a block matrix to facilitate the prior specification and design a MCMC sampling scheme.
|
