Speed up post-clustering processing about hs2

[frozenblit]

Have you tried using spikes.groupby('cl')? Then you can loop over the tuples (cluster_number, spikes within that cluster) and use all the pandas machinery to efficiently calculate the grouped means and amplitudes. In my experience that was much faster than iterating over the cluster labels.

[mhhennig]

Indeed, wow, thank you Fernando! This is now much faster, and as a side effect, DBSCAN now works flawlessly.

[frozenblit]

Pandas is magical ;)

[martinosorb]

Ah, this is my fault, I don't understand Pandas very well. Thanks Fernando!