Figure 1. Generation of the synthetic recordings. A, A total of 20 cells are randomly placed in front of a 32-channel NeuroNexus probe layout. The plot shows the location of each cell for one recording. B, Sample template pairs generated by neurons with different cosine similarity values. C, Cosine similarity matrix between all pairs of templates for a sample recording. D, Cosine similarity as function of the distance between the neurons, either using the real position from the simulations (orange circles), or the estimated barycenter of the templates (blue circles). E, Histogram of the cosine similarity distribution from one of the simulated recordings. F, Cross-correlograms and auto-correlograms for three sample spike trains. G, Average auto-correlograms of all units (red line, gray area represents the SD). H, Average cross-correlogram over all pairs of neurons (red line, gray area represents the SD around the mean). I, Sample traces from 10 channels of one synthetic recording.