Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Making Sense of Dependence: Efficient Black-box Explanations Using Dependence Measure

About

This paper presents a new efficient black-box attribution method based on Hilbert-Schmidt Independence Criterion (HSIC), a dependence measure based on Reproducing Kernel Hilbert Spaces (RKHS). HSIC measures the dependence between regions of an input image and the output of a model based on kernel embeddings of distributions. It thus provides explanations enriched by RKHS representation capabilities. HSIC can be estimated very efficiently, significantly reducing the computational cost compared to other black-box attribution methods. Our experiments show that HSIC is up to 8 times faster than the previous best black-box attribution methods while being as faithful. Indeed, we improve or match the state-of-the-art of both black-box and white-box attribution methods for several fidelity metrics on Imagenet with various recent model architectures. Importantly, we show that these advances can be transposed to efficiently and faithfully explain object detection models such as YOLOv4. Finally, we extend the traditional attribution methods by proposing a new kernel enabling an ANOVA-like orthogonal decomposition of importance scores based on HSIC, allowing us to evaluate not only the importance of each image patch but also the importance of their pairwise interactions. Our implementation is available at https://github.com/paulnovello/HSIC-Attribution-Method.

Paul Novello, Thomas Fel, David Vigouroux• 2022

Related benchmarks

TaskDatasetResultRank
ExplainabilityImageNet (val)
Insertion48.1
104
Attribution FidelityImageNet 1,000 images (val)
µFidelity0.202
48
Discovering the causes of incorrect predictionsCUB-200-2011 (val)
Avg Highest Confidence (Top 25%)0.2418
31
Audio Classification AttributionVGG-Sound (val)
Deletion AUC9.04
28
Attribution FaithfulnessImageNet (val)
Deletion8.66
22
Attributing Multimodal Foundation Model ErrorsImageNet misclassified samples (val)
Avg Highest Confidence (0-25%)21.48
22
AttributionImageNet (val)--
20
Attribution FaithfulnessImageNet CLIP ViT-L (val)
Deletion Score0.1565
12
Attribution FaithfulnessImageNet ImageBind Huge (val)
Deletion Score0.1875
11
Attribution FaithfulnessImageNet LanguageBind Large (val)
Deletion Score0.1034
10
Showing 10 of 16 rows

Other info

Code

Follow for update