Glm pynapple nemos tutorial #451
Conversation
…ull of nans at event presentation)
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
…plots are done, fully functional.
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:09Z What is the criteria for responsiveness? That seems like an important aspect to briefly explain? camila-maura commented on 2025-07-10T03:33:45Z I hid the function for it because it is indeed just a normalized difference but if people are keen on looking into it they can click on the funcition and see it - I believe it is fairly well commented (: |
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:10Z duplicated with previous sentences camila-maura commented on 2025-07-10T16:28:27Z done |
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:11Z "imply assuming" camila-maura commented on 2025-07-10T03:36:09Z done
|
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:12Z typo at "which is a makes sense"
camila-maura commented on 2025-07-10T20:43:40Z done |
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:12Z So the coupling term is the averaged of all other neurons? It is not clear to me how this line : nmo.basis.RaisedCosineLogConv(
Does this everaging? Maybe add a few sentences? camila-maura commented on 2025-07-10T16:48:25Z Added an admonition explaining and changed the text a little bit. the definition of the basis creates the object, and then by calling compute_features we convolve the raised cosine log conv with the input (the spike counts of all units). |
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:13Z Very nice! camila-maura commented on 2025-07-10T16:48:36Z (: |
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:13Z Order of ideas is a bit odd. I think you meant to say comparing values are valuable within datasets but need to be normalized when comparing across datasets. Maybe just re-arrange the sentences for clarity? camila-maura commented on 2025-07-10T16:50:58Z I removed the log likelihood and just kept the scoring with pseudo r2 |
|
View / edit / reply to this conversation on ReviewNB jeromelecoq commented on 2025-07-08T19:34:14Z Very nice addition! camila-maura commented on 2025-07-10T16:48:47Z (: |
|
View / edit / reply to this conversation on ReviewNB rcpeene commented on 2025-07-08T20:40:45Z Could you briefly cite/mention the dataset being used and why? Additionally just briefly mention you're using nmo's download function which is different than the rest of the notebooks. camila-maura commented on 2025-07-10T16:50:07Z Full data citation is at the bottom of the notebook (: Added a comment before the call to the download function camila-maura commented on 2025-07-10T20:47:45Z (we are using the same dataset as you!) |
…+ raster of subset of units moving forward.
|
I hid the function for it because it is indeed just a normalized difference but if people are keen on looking into it they can click on the funcition and see it - I believe it is fairly well commented (: View entire conversation on ReviewNB |
|
done View entire conversation on ReviewNB |
|
done
View entire conversation on ReviewNB |
|
done
View entire conversation on ReviewNB |
…revision of prediction nan outputs at end and beginning
|
done View entire conversation on ReviewNB |
|
Added an admonition explaining and changed the text a little bit. the definition of the basis creates the object, and then by calling compute_features we convolve the raised cosine log conv with the input (the spike counts of all units). View entire conversation on ReviewNB |
|
(: View entire conversation on ReviewNB |
|
(: View entire conversation on ReviewNB |
|
Full data citation is at the bottom of the notebook (: Added a comment before the call to the download function View entire conversation on ReviewNB |
|
I removed the log likelihood and just kept the scoring with pseudo r2 View entire conversation on ReviewNB |
|
done View entire conversation on ReviewNB |
|
(we are using the same dataset as you!) View entire conversation on ReviewNB |
|
We chose a smaller bin size to better capture transient dynamics in some of our units - using larger bins tended to result in poorer predictions. The smoothing applied in the plots is purely for visualization: the plots still look OK without the smoothing, but appear noisier, which can be distracting. View entire conversation on ReviewNB |
2 participants
Jupyter notebook for modeling spiking neural data with GLMs, using Pynapple and NeMos python packages.
(Internal review until undrafted)