Extended Data Fig. 7: Rastermap embedding quality as a function of parameters.

In Fig. 1, we ran Rastermap with 100 clusters (k-means clustering), locality parameter of 0.8, and time_lag_window of 10. Here we run Rastermap with varied parameter values. a, Rastermap was run with different numbers of clusters (black) for each of the ten simulations from Fig. 1 (error bars represent s.e.m.). We also ran Rastermap using the Leiden algorithm to perform clustering with a resolution of 3.0 and 100 neighbors, which produced 100 clusters (gray) – this clustering method performed worse than k-means. Percent correct triplets, percent contamination and chance level (dashed line) computed as in Fig. 1i,j. b, Average percent correct triplets and percent contamination computed from Rastermap sorting from the ten simulations, using different locality and time_lag_window values.