The main parameters used by Spikes_Link are:
- \(N_{min}\): minimum number of spikes recommended for a given block (it could depend on the spike sorting algorithm chosen).
- \(N_{max}\): maximum number of spikes recommended for a given block.
- \(T_{max}\): maximum period of time where the recording can be presumed stable.
- \(N_{OS}\): number of samples per class in the overlapping set.
- \(p_{out}\): proportion of spikes that would be discarded as potential outliers to create the overlapping set.
- \(MS\): number of consecutive blocks for a temporal merger before it is confirmed.
- \(CS\): number of consecutive blocks that a class can survive without a matched cluster. Note the constraint \(MS \leq CS\).
- \(S_{TH}\): minimum amount of blocks with spikes for a non-spurious class.
Some parameters can depend on the specific spike sorting algorithm of choice (\(N_{min}\), \(N_{max}\)) or the expected quality of the clustering (\(p_{out}\)). An initial expected stability is parameterized by \(T_{max}\), with some drifting being allowed within a block without compromising the performance (as seen in Fig. \ref{582093}), although this might depend on the chosen spike sorting algorithm. In this work, \(N_{OS}\) was set to 500; a much larger value would increase the computing time used by the spike sorting algorithm, whereas a much smaller value would affect the performance of the overlapping metric. For sparse neurons that could be absent in some blocks, the parameter \(CS\) will cap the number of blocks in which the same waveform will be searched. \(MS\) measures how many mergers of previous classes need to be detected before loss of isolation is confirmed. This is also important, as blocks with high levels of noise can affect the performance on a given block, and having \(MS>1\) gives the chance to the algorithm to continue isolating individual clusters in the following blocks. Finally, \(S_{TH}\) defines the minimum scale in blocks at which a class is considered as valuable. Therefore, the parameters are very intuitive and have a direct relationship with the experimental conditions. Moreover, it is not necessary to define other thresholds like the maximum distance between clusters, which is hard to estimate but commonly used in tracking algorithms.