Fig. 5: Simulation of trial-by-trial variability controlled by input covariance in a PV-dominated regime.

a The input covariance matrix for sampling input point clouds (b). From i to ii (iii), the trial-by-trial input covariance between E and SST (PV) populations were increased. In column iv, covariances between E-P and E-S pairs were both increased. a bottom Output covariance matrix across 100 trials where inputs for each trial were sampled from the covariance matrix in a top. The output variance of the excitatory population is indicated with a red square. b Trial-by-trial input point clouds sampled from given covariance matrix as columns i to iv in a top. (axes are normalized to the maximal value in Fig. 1d) c Peristimulus time histogram (PSTH) of excitatory population response. Black lines: individual trial. Red line: average response over 100 trials. Stimulus (which involved only a change in the variance and/or covariance without any change in the mean) was provided at 250 ms, and the output covariance matrix in the a bottom was calculated for the last 500 ms.