Example of different scenarios of graph reduction on 11 consecutive blocks. A For a given class, multiple nodes are created, one for each block from the creation of the class until its possible disappearance. Filled circles are nodes associated to classes that were updated on that block with new spikes. The edges of the graph connect the nodes across different blocks according to the rules explained above in the tracking loop. Empty circles with a solid line denote classes that were not updated with new spikes on that block, although their overlapping sets were included in the sorter. Different classes are represented by different colors. The arrows at the very left of the graph denote that the class has been tracked over several previous blocks, a class with no arrow on the left is a new class, and if a class ends with an arrow at the very right of the graph it means that it will continue being tracked afterward. Sub-graph f has no stable classes so it will be discarded. In the blue node is associated to a new class at \(t_i\) and matched in the next block; next, it merged with the stable orange class, leading to the green nodes, with the merge being confirmed after 3 blocks. This behavior could be the result of overclustering (blue class) by the sorting algorithm. In the green class has a temporary overclustering (red class), that was not automatically merged in the following block, but was merged in the next one (light green node), but only for that block. In c, multiple SC (grey and brown) are created (possibly due to overclustering), but they are not all merged together on the next block. Only partial temporal merges with other classes were found, without sharing the exact same ML list. The case d depicts a temporary merge during only one block, which could be easily caused by changes in the noise level of the recording. After a few blocks of separable activity, the classes were then combined in a confirmed merger. Resulting graphs following the graph reduction criteria. In this case, 3 different classes are obtained during the 11 blocks of recording analyzed. Case uses criterion 2, whereas b uses criterion 1; both, in turn, are combined into a single class following merger reduction. Notice that the class was not found during block \(t_{i+9}\) (dotted circle), but the survival criterion prevented it from disappearing. Case c is combined all the classes into the yellow using criterion 2, and using merge reduction in conjunction with case d, leads to 2 classes being tracked.