Comparison of metrics for example spike trains.

A) Sensitivity to spike timing and spike count for different measures for an example of three synthetic spike trains (as in [32] (Fig 5D)). Bottom: Computation of Victor-Purpura metric for pairs of spike trains, varying the shift-cost q (left), and computation of distances for SPIKE, RI-SPIKE, and SpikeShip (right) (see Methods). SpikeShip and RI-SPIKE dissimilarities reflect spike timing rather than spike count. By contrast, VP reflects spike-count differences, which holds true for the entire range of q. B) Sensitivity of different measures to global shifts of spiking patterns and changes in relative spike timing. Top: Example of three multi-neuron spiking patterns. Onset of spike patterns B and C where shifted relative to pattern A. Colors correspond to the shift/delay applied to half (blue) and the full neural population (green). Bottom: Comparison between spike train dissimilarities between patterns (A, B) (blue) and pattern (A, C) (green).