jmarkow / spikoclust

SPIKOsaurus CLUST: A MATLAB toolbox for split-and-merge-expectation-maximization-based spike sorting
Other
1 stars 1 forks source link

Giving error 'not all covariance entries are finite', unclear how to fix #3

Open danielpollak opened 6 years ago

danielpollak commented 6 years ago

For some data sets, the analysis will fail with a 'not all covariance entries are finite' error in spikoclust_gmem_covfix.m. It is unclear what if anything can be done to remedy this, nor is it clear where this error originates from.

jmarkow commented 6 years ago

Are you running the fully automated sorter or doing things stepwise per http://spikoclust.readthedocs.io/en/latest/usage.html#stepwise-sorting ? It's possible something is getting borked between spike detection/signal conditioning and the clustering step. I would check to make sure the spikes look reasonable before proceeding with clustering.

danielpollak commented 6 years ago

I was running the fully automated sorter. I can do stepwise though. Would you recommend doing anything besides changing around the filter?

Also, can you give me an idea of what the error I got means?