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

Add a feature filter before feature selection #394

Open
wants to merge 5 commits into
base: master
Choose a base branch
from

Conversation

gvling
Copy link

@gvling gvling commented Oct 30, 2019

I just add a filter before selection features.
When I use LIME to get an explanation with a model, I wouldn't get some features are come out in explanation result.
So I implement a filter before the feature selection.
That also can decrease the feature selection's search space to speed up the feature selection.

@gvling gvling changed the title Add a feature filter before feature selection WIP: Add a feature filter before feature selection Oct 30, 2019
@gvling gvling changed the title WIP: Add a feature filter before feature selection Add a feature filter before feature selection Oct 30, 2019
@marcotcr
Copy link
Owner

What is the use case for this? The model is using features 1-10, but you are only interested in explanations for features 1-5?
I have some comments on the code, but first I am trying to understand why we need this option.
Thanks,

@gvling
Copy link
Author

gvling commented Dec 23, 2019

When I use LIME to explain some instances, the result of explanations is changed randomly.
I know that there have two random elements in LIME(features, and samples).
But in some use cases, we roughly know which features impact the result stronger, or we just interested in explanations for features 1-5:

  • In the Titanic data set, we just interested in features exclude gender, name and don't want to change the original model(because it may change the performance of model).

In the above cases, selectable features may useful.
Thank you.

@marcotcr
Copy link
Owner

marcotcr commented Apr 3, 2020

I don't think I am convinced that this is a feature we need - would anyone be interested in excluding certain features, even if they are important to the model?

@tigermli
Copy link

Interested. This would be very useful in situations where each example only depends on a small subset of the features in a dataframe and considering all of the features takes a while.

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.

3 participants