Add models with no-ic [training and LOIO] #51
Conversation
* update download module * rename saved labeled data * update split data module * add new train models script * update readmes * update running model code * update get x y function * mike suggestions * add new models
* update evaluate module * update scripts * add documentation * documentation * remove unecessary notebook * Update 3.evaluate_model/README.md Co-authored-by: Gregory Way <gregory.way@gmail.com> * add comment --------- Co-authored-by: Gregory Way <gregory.way@gmail.com>
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
|
||
| ### Data Preview | ||
|
|
||
| The labeled dataset includes CellProfiler (CP) and DeepProfiler (DP) features as well as metadata (location, perturbation, etc) for cells from the original MitoCheck project. | ||
| The breakdown of cell counts by phenotypic class (as labeled manually by MitoCheck) is as follows: | ||
| The breakdown of cell counts by manually-assigned phenotypic class for the `ic` (illumination corrected) dataset is as follows: |
There was a problem hiding this comment.
why is it necessary to specify for the IC dataset? Isn't it the same for no-ic?
There was a problem hiding this comment.
I believe they differ (very slightly). Because illumination correction has a very small effect on segmentation, we get different cell locations and in some cases a different number of cells.
I was going to update this when I update the path that we download the no_ic data from in WayScience/mitocheck_data#37. See below for more explanation on this
There was a problem hiding this comment.
Just one comment to address, then LGTM!
Your plan for merging/addressing comments sounds good. I would recommend waiting to merge this until after we merge WayScience/mitocheck_data#37 and integrate the new download link so we don't have an unclear download path on main.
Sounds good, thanks @roshankern ! Please give me a 👍 when I can merge. (I will also fix your file name comment ASAP). |
Will do! |
|
#37 looks great @roshankern, I just reviewed. I think there is one small change to do. I also addressed the file name fix you suggested above. Once you merge in #37 feel free to merge this one. |
@roshankern can you review my contributions (the last two commits), and I will review your contributions (first two commits).
My plan is to get this merged in, address comments in #47, merge #47, and then start a new visualization with a clean repo.
I looked at the results briefly, and they don't seem to be changed much (i.e. IC doesn't seem to have a big impact in this case).