MouseLand
Renaming Kilosort variables for readability
I have been reading through the Kilosort4 source code to better understand how it works, and I realize that less opaque variable names (e.g., tF, st, Wall, clu) would make this a lot easier. For example,
tF --> local_pca_feats st --> spike_times clu --> spike_clusters NT --> batch_size nt --> waveform_length Wall --> cluster_channel_pcs
If I were to undertake making these changes over an extended period (in my own fork), is this something that the maintainers of Kilosort4 would be interested in incorporating into the project?
We're already planning to make these kinds of changes, it just hasn't happened yet. I have to coordinate with @marius10p when deciding on new naming conventions, so it's likely we'll want to decide on those changes ourselves. Some of the variable and parameter names are intentionally unchanged to keep them consistent with the paper and past versions of Kilosort, for example.
Understood. In that case, some sort of key mapping the original variable names to what they signify could be a useful interim step!
That's what comments/documentation are intended to accomplish, we're just behind on updating that. I will note, however, that most of the variables you list are in fact documented in kilosort.run_kilosort.
I'm going to close this for now, but I'll send another reminder about working on adding more comments to the source code.