Skip to content

Conversation

@chrisfosterelli
Copy link
Contributor

@chrisfosterelli chrisfosterelli commented May 29, 2017

Reference Issue

Fixes #7955

What does this implement/fix? Explain your changes.

This proposal changes the description of the RBF SVM parameters example to not include confusing wording relating to C limiting the number of support vectors in the wrong direction, and instead focuses on a more interpretable description that relates to practical usage.

Any other comments?

I need some feedback that this description is intuitive and is clear to understand, and actually an improvement in the right direction :)

Specifically:

  1. Is "decision function" the right word here when referring to margin?
  2. We don't include any explanation or example or visualization of the margin. Can we expect the audience of this to know what that is? The visualization here captures it great but I don't know if we can include that image in the documentation since it's originally authored by someone else.

Copy link
Member

@jnothman jnothman left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is an improvement

against maximization of the decision function's margin. For larger values of
``C``, a smaller margin will be accepted if the decision function is better at
classifying all training points correctly. A lower ``C`` will encourage a larger
margin, and a more simple decision function, even if that decision function
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"and a more simple decision function [...]." -> "therefore a simpler decision function, at the cost of training accuracy?

Without the "therefore" it might seem that "simplicity" is a different measure that we're not defining, (and frankly would be tricky to define.) This way we just suggest that the two are related.

The other part of my proposed change is simply reducing verbosity :)

use phrasing suggested by @vene.
@amueller amueller merged commit 974ceb5 into scikit-learn:master May 23, 2018
@chrisfosterelli
Copy link
Contributor Author

chrisfosterelli commented May 23, 2018 via email

@amueller
Copy link
Member

Only a year or so ;)

qinhanmin2014 added a commit that referenced this pull request May 24, 2018
I think there's concensus on not to introduce new flake8 errors
cc @amueller
@chrisfosterelli chrisfosterelli deleted the fix-plot-rbf-params-docs branch April 1, 2019 20:52
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.

SVM documentation clarification: Higher values of C -> more support vectors?

4 participants