Example of results of the simulated recording number 21. The mean waveforms of the classes and the peak to peak amplitude of their spikes are shown for the ground truth (top) and each algorithm. The dotted vertical lines delimit the segments that are used by Spikes_Link as blocks; the mean waveforms for each class are also calculated within these segments. For each algorithm we computed the mean value of the accuracy (\(\mu_a\)), recall (\(\mu_r\)) and precision (\(\mu_p\)), and their range (\([min,max]\)) across all the simulated classes. Accuracy of each algorithm; Spikes_Link_WC: \(\mu_a\)=0.998, [0.99, 1]; Wave_clus: \(\mu_a\)=0.52, [0.26, 0.83]; Combinato: \(\mu_a\)=0.49, [0.21, 0.73]. Recall of each algorithm; Spikes_Link_WC: \(\mu_r\)=0.99, [0.996, 1]; Wave_clus: \(\mu_r\)=0.53, [0.26, 0..83]; Combinato: \(\mu_r\)=0.66, [0.5, 0.76]. Precision of each algorithm; Spikes_Link_WC: \(\mu_p\)=0.99, [0.997, 1]; Wave_clus: \(\mu_p\)=0.99, [0.97, 1]; Combinato: \(\mu_p\)=0.75, [0.23, 1]. In this example, Spikes_Link_WC tracked the drifting classes correctly, Wave_clus generated a strong overclustering, and Combinato also presented overclustering but merging two neurons together as well. The other metrics are shown in Fig. \ref{129092}.