A1308
Title: Modelling neural connectivity with Bernoulli autoregressive processes
Authors: Carolina Euan - Lancaster University (United Kingdom) [presenting]
Abstract: Understanding spiking neuron interactions is vital for deciphering cognitive processes like memory and decision-making. A novel approach using the Bernoulli Autoregressive Partially Linear Additive (BAPLA) model is presented to infer connectivity from nonstationary, high-dimensional spike train data. The method utilizes a regularized maximum likelihood estimator with an $l_1$ penalty for sparse and interpretable neuronal interaction estimates. Nonstationarity in firing rates is addressed by incorporating a nonparametric trend and implementing a robust inference procedure to quantify uncertainty in the findings.
