EIG did not converge (for Utah array data)

[liz-ok]

We are working with a 96ch Utah array data. It runs without errors when using a linear widely spaced channel map. However, when rearranging the channel map to represent a 2D utah array (attached), we run into this issue after main optimization:

Error running kilosort! EIG did not converge at index = 71926

followed by:

Error saving data for phy! Unrecognized filed name "Wphy".

Windows 11
Matlab R2022a
Cuda 12.1
UEA_kilosort_chanmap.zip

Any ideas?

Thank you!!

[kermit000000]

I've run into a similar issue before, and I still can't figure out what aspect of the channel map triggers an error in the final step of KiloSort. But from my experience, it seems that arranging the xcoords and ycoords in incremental order (or linear order) in the channel map resolves the problem. On the flip side, when the chanMap numbers are in sequence and the corresponding xcoords and ycoords are all jumbled up, it throws an error.

While I can't guarantee the absolute truth of this observation, I recommend giving it a shot—organize your xcoords and ycoords in order while mixing up the corresponding chanMap and chanMap0ind numbers. For instance, your current map starts chanMap 1 at coordinates (800,3600), then 2 at (400,3600). However, try the sequence with chanMap number 79 at (0,400), followed by 77 at (0,800), 75 at (0,1200), and so on. It worked for me; hopefully, it does the trick for you too.