Update who feature selection example

scikit-learn-contrib / scikit-matter

A collection of scikit-learn compatible utilities that implement methods born out of the materials science and chemistry communities

https://scikit-matter.readthedocs.io/en/v0.2.0/

BSD 3-Clause "New" or "Revised" License

70 stars 18 forks source link

Update who feature selection example #212

Closed agoscinski closed 10 months ago

agoscinski commented 10 months ago

The plot has been updated with the revision of the ore paper. See corresponding plot in the ore repo https://github.com/victorprincipe/skmatter-ore/blob/main/paper/WhoDataset-Selection.ipynb

change kernel ridge to linear regression
add correlation plot that shows similarity between the selector methods

:books: Documentation preview :books:: https://scikit-matter--212.org.readthedocs.build/en/212/

agoscinski commented 10 months ago

this added plot at the end of this example https://scikit-matter--212.org.readthedocs.build/en/212/examples/selection/FeatureSelection-WHODataset.html#sphx-glr-examples-selection-featureselection-whodataset-py is the main change

ceriottm commented 10 months ago

Observation for both the notebook and the paper: I would only show the lower-left corner, and also not show the diagonal. Those are redundant.

agoscinski commented 10 months ago

Updated it in the paper but I would keep this redundancy for the example. Redundancy can help with understanding. Also making such a plot with matplotlib would make the code more complicated which also reduces readability.

ceriottm commented 10 months ago

Updated it in the paper but I would could this redundancy for the example. Redundancy can help with understanding. Also making such a plot with matplotlib would make the code more complicated which also reduces readability.

OK I see the point of keeping the code clean in the example (although in many examples we have pretty sophisticated matplotlib layouts). Good with me then