log a warning if binary classifier SHAP tensors do not add up to 0, instead of raising an AssertionError #24
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
… instead of raising an AssertionError
j-ittner
added
the
sequential
j-ittner
requested review from
jason-bentley,
mgelsm,
konst-int-i and
joerg-schneider
jason-bentley
approved these changes
Aug 25, 2020
j-ittner
deleted the
bugfix/raw_shap_assertion_for_binary_classification
branch
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Sequential PR: Please first approve and merge PR #6 before processing this PR.
This resolves #12.
By definition, both SHAP tensors obtained for a binary classifier should add up to 0.0 for each observation and feature.
In practice we have seen evidence of totals up to 0.01 due to imprecisions in the SHAP explainer's approach for estimating SHAP values.
Therefore we will not raise an exception if the totals are not 0.0 but will log a warning instead, stating the range of observed totals. As long as these totals are small (e.g., less than 0.05, corresponding to 5%pt probability), it should be safe to ignore these warnings.