Dissimilarity matrices and clustering comparison.

A) Six multi-neuron spiking patterns that were generated according to different inhomogenous Poisson processes, plus one noise pattern (homogeneous Poisson). We show the first 50 neurons from a total of N = 500 neurons. The first 180 epochs correspond to homogenous noise (random firing), and the next 180 correspond to pattern that were generated based on an inhomogenous Poisson process (6 different patterns with 30 epochs each). Each pattern was simulated with Tepoch = 300 samples, Tpulse = 30 samples. The firing rates outside the pulse were λout = 0.02 spks/samples and during the pulse λin = 0.2 spks/samples, respectively. The homogenous noise pattern had the same average firing rate as the inhomogeneous patterns. B) Sorted Dissimilarity matrix by pattern using SpikeShip dissimilarity measure (left), and a 2-dimensional t-SNE embedding using SpikeShip dissimilarity matrix (right).