Graph Reduction

After the analysis of a long-term recording has been performed, some classes will be present in only a few blocks, and later disappear or be merged with others. Usually, these classes are non-resolved overclusterings from another class within a block, and in most cases, they should be combined together. At the same time, some of them can be created by errors from the spike sorting algorithm and should be removed to preserve the quality of the results. We define as spurious classes (SC) those that appeared in less than \(S_{TH}=3\) blocks throughout the whole recording, and the clusters associated to the classes in those blocks are not labeled as mergers. Otherwise, if a class appears in at least blocks \(S_{TH}\), it is considered a "stable class".