These are chat archives for epnev/ca_source_extraction

25th
Oct 2016
Gilles Vanwalleghem
@Yassum
Oct 25 2016 06:26
refine image.png
@epnev Hi, I am having a weird bug with manually_refine_components. The display is fairly garbled with a purple overlay until I click on something and then it clears up for a second before going back to strange looking
Here is what it looks like when "clean"
ROIs image.png
eftychios pnevmatikakis
@epnev
Oct 25 2016 12:06
@Yassum Not sure.. are you also passing in the background b when calling manually_refine_components? How many components do you have? Is this somatic or dendritic imaging? This function will only work for somatic imaging
pwang724
@pwang724
Oct 25 2016 16:31
Hi @epnev, I've been using your software for quite a while now and it's been great. I've recently started to play with your deconv methods and based on the experience of people here, MCMC seems to outperform CD. However, when I attempt MCMC, as compared to constrained foopsi, the resulting traces appear wildly divergent from the raw data, while the foopsi appears fine (as shown here: https://postimg.org/image/hfagl4vrp/). I may have to ask you in more in-depth and provide you with code, but are there some parameters I could tweak to attempt to fix this issue? Thanks!
eftychios pnevmatikakis
@epnev
Oct 25 2016 19:00
Hi @pwang724
Yes, it's hard to tell from this image what might be wrong. Best if you send me some traces where the MCMC method fails.