search

csn-le / wave_clus

A fast and unsupervised algorithm for spike detection and sorting using wavelets and super-paramagnetic clustering
121 stars 65 forks source link

Problem with S probe recording #186

Closed aboharbf closed 2 years ago

aboharbf commented 3 years ago

Running into an issue w/ S probe recordings. 32 channels, and each one I've checked yields: """

SPC result of file: 20201115Mo001_NSX_Ch1_spikes

SPC v2.1 (14/2/2000) -- efficient KNN and mstree
SPC v2.0.1 (1/12/2000) -- StopRunAtBreak option
SPC v2.0 (1/11/1999) -- new param file + revised data structures

error: at line 233 of 'param.c': N<=0 """

Any tips or trouble shooting advice would be great.

ferchaure commented 3 years ago

try calling get_spikes first and check the spikes matrix in the results, something could be wrong with the input to the SPC algorithm. Other common issue is having another library on path that is overwriting some wave_clus function.

aboharbf commented 3 years ago

So I added a try loop, and it seems like 4 channels specifically are causing the issue - 1, 5, 26, and 32. Looking at each _spike file, nothing clear seperates channels making this errors and ones that do: image tmp is channel1, which failed, while tmp2, channel 2, has way fewer spikes but did not fail. The 32 channels are processed together, so I'd be surprised if it was a path problem, especially given that seemingly random ones in the middle fail as well. Anything else I may be able to try?

ferchaure commented 3 years ago

an option is to plot the spikes and see if they are spikes or maybe a strong noise or acquisition error.

another stronger check is to add a breakpoint in line 548 of Do_Clustering an see the values in inspk_aux.