Skip to content

Fix Interpretability Methods target_class_idx#926

Open
Logiquo wants to merge 16 commits intosunlabuiuc:masterfrom
Logiquo:fix/interpret_bin
Open

Fix Interpretability Methods target_class_idx#926
Logiquo wants to merge 16 commits intosunlabuiuc:masterfrom
Logiquo:fix/interpret_bin

Conversation

@Logiquo
Copy link
Copy Markdown
Collaborator

@Logiquo Logiquo commented Mar 31, 2026

  1. Unify all interpretability to support target_class_idx and its semantics
  2. target_class_idx becomes a no-op for binary classification, and select target class we want to interpret for multi-class and multi-label.
  3. The multi-label will be selected based on argmax if target_class_idx is not provided.
  4. All samples (including negatives) will be evaluated in the interpretability metric, can support sample_filter which can specify POSITIVE, NEGATIVE and IGNORE based on the predicated probablity.

@Logiquo Logiquo requested a review from jhnwu3 April 1, 2026 00:26
@Logiquo Logiquo marked this pull request as draft April 7, 2026 07:36
Copy link
Copy Markdown
Collaborator

@jhnwu3 jhnwu3 left a comment

Choose a reason for hiding this comment

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

Just wondering if it's possible to get more examples of running this new modification here.

@Logiquo Logiquo marked this pull request as ready for review April 7, 2026 23:27
@Logiquo
Copy link
Copy Markdown
Collaborator Author

Logiquo commented Apr 8, 2026

image image

We can have custom function to determine what sample is considered as positive/negative/ignore.

@Logiquo Logiquo requested a review from jhnwu3 April 8, 2026 01:18
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.

2 participants