Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[ENH] OWTestLearners: Cross validation by feature #2145

Merged
merged 2 commits into from
Apr 7, 2017

Conversation

VesnaT
Copy link
Contributor

@VesnaT VesnaT commented Mar 28, 2017

Issue

Add type of CV where number of feature values determines number of folds.
Use only discrete meta features.

Description of changes

Add radio button option to select this type of CV.
Add combo box to select feature to give nr. of folds.
Documentation comes when gui is approved.

Includes
  • Code changes
  • Tests
  • Documentation

@VesnaT VesnaT changed the title OWTestLearners: Cross validation by feature [ENH] OWTestLearners: Cross validation by feature Mar 28, 2017
@VesnaT VesnaT force-pushed the marko-cv_feature branch from bb4c530 to e5edb48 Compare March 28, 2017 12:10
@codecov-io
Copy link

codecov-io commented Mar 28, 2017

Codecov Report

❗ No coverage uploaded for pull request base (master@51c97b6). Click here to learn what that means.
The diff coverage is 89.28%.

@@            Coverage Diff            @@
##             master    #2145   +/-   ##
=========================================
  Coverage          ?   72.08%           
=========================================
  Files             ?      318           
  Lines             ?    54653           
  Branches          ?        0           
=========================================
  Hits              ?    39397           
  Misses            ?    15256           
  Partials          ?        0

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 51c97b6...2a5bf2f. Read the comment docs.

@VesnaT VesnaT changed the title [ENH] OWTestLearners: Cross validation by feature [WIP][ENH] OWTestLearners: Cross validation by feature Mar 28, 2017
@VesnaT VesnaT changed the title [WIP][ENH] OWTestLearners: Cross validation by feature [ENH] OWTestLearners: Cross validation by feature Mar 29, 2017
@VesnaT VesnaT force-pushed the marko-cv_feature branch from a9c5b5c to a2802d6 Compare March 29, 2017 08:18
@VesnaT VesnaT force-pushed the marko-cv_feature branch from a2802d6 to e12109d Compare March 31, 2017 08:31
@janezd
Copy link
Contributor

janezd commented Mar 31, 2017

I know this will come as a totally unusual surprise, but I'm disapproving the GUI. :)

The feature, if I read the code correctly, doesn't determine the number of folds, but the folds themselves. This looks like a useful feature, but I wouldn't expect it in cross-validation, and particularly not in the "Number of folds" combo. "From feature" also doesn't fit into an unexpanded combo.

The "from feature" option fundamentally changes the sampling procedure in the sense that it is no longer random.

As a minor argument against this GUI, we also discourage show/hide, and prefer enabling/disabling instead, where this could be sensibly avoided.

One solution would be to add a new radio button "Cross validation by feature" below "Cross validation". With this, this functionality would be more prominent and we'd no longer need hiding/showing of parts of the interface.

The radio button can be disable if there are no meta attributes.

If you do this, don't forget to provide appropriate migration of settings.

@VesnaT VesnaT force-pushed the marko-cv_feature branch from e12109d to 6ca4722 Compare April 3, 2017 10:16
@VesnaT VesnaT added this to the 3.4.2 milestone Apr 7, 2017
@markotoplak
Copy link
Member

The main functionality is not tested (the corresponding radio button is never clicked). :)

When cross validation by feature is used for a file, which is then disconnected and connected back, the setting are reset.

@VesnaT VesnaT force-pushed the marko-cv_feature branch from 6ca4722 to e01b3ca Compare April 7, 2017 10:31
@VesnaT VesnaT force-pushed the marko-cv_feature branch from e01b3ca to 2a5bf2f Compare April 7, 2017 11:25
@markotoplak markotoplak merged commit 9d6de2c into biolab:master Apr 7, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants