- Notifications
You must be signed in to change notification settings - Fork3.3k
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 ourterms of service andprivacy statement. We’ll occasionally send you account related emails.
Already on GitHub?Sign in to your account
Add Interaction Order to Explanation Class#3939
base:master
Are you sure you want to change the base?
Conversation
tests/plots/test_image.py Outdated
| ) | ||
| # define the model architecture | ||
| inputs = tf.keras.Input(shape=(32, 32, 3)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
we have a problem here. Retraining this won't reproduce the image. Any ideas how we can get this to work? I would like to reproduce the issue from the ticket
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others.Learn more.
I don't understand why the tests fail. Just from the files that differ it looks like my test has something to do with it but when looking at the diffs I cannot understand why. I also cannot reproduce the issue locally.
Okay, just found that exactly these tests fail in other pipelines as well. Will try to reproduce and create a fix |
I think this was closed in the mean time |
is it really closed, take a deeper look |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@## master #3939 +/- ##==========================================+ Coverage 64.78% 64.80% +0.01%========================================== Files 92 92 Lines 12895 12905 +10 ==========================================+ Hits 8354 8363 +9- Misses 4541 4542 +1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
Note that this PR solves the issue only for the |
Overview
Closes#3874
Description of the changes proposed in this pull request:
base_valuesnoroutput_nameswas provided when creatingExplanationinstances from the DeepExplainer which resulted in an empty tuple asExplanation.output_dims, which then caused problems in theimage_plot. We solve this here in two different fashions: 1. set the base_values correctly, 2. add interaction_order. Each of these solutions would be fine by itself, but I think, the interaction_order can help to track down if everything was done correctly and helps to understand theExplanation's shap_values better.interaction_orderto the explanation class. This is either0for no interactions or1if interactions are present. I also added this to a couple of tests.Checklist