ShapBPT in Perspective: A Consolidated Review and an eXplainable Anomaly Detection Case Study
Published in QualITA Workshop | ICPE 2026 (ACM), 2026
This paper presents ShapBPT in Perspective, a consolidated review and practical case study of ShapBPT for eXplainable Anomaly Detection (XAD). The work bridges hierarchical Shapley-based explanations with real-world anomaly detection systems, demonstrating how structured feature attributions can improve the interpretability of black-box models.
Unlike traditional explanation methods based on fixed partitions, ShapBPT leverages a data-aware Binary Partition Tree (BPT) to generate multiscale, semantically aligned explanations. This enables more faithful localization of anomalous regions and provides clearer insight into model behavior.
🔗 Resources
- Paper: ACM Digital Library
- Code: XAD GitHub Repository
- Workshop: QualITA Workshop
- Conference: ICPE 2026
📌 Key Contributions
- A consolidated review of ShapBPT for explainable computer vision.
- A real-world case study on explainable anomaly detection (XAD).
- A unified framework connecting hierarchical Shapley explanations with anomaly detection systems.
- Release of open-source implementation (XAD) for reproducibility.
⚙️ Method Overview
ShapBPT explains anomaly detection decisions by assigning attribution scores to image regions using a hierarchical, data-driven partitioning strategy.
Example Explanation
XAD Workflow
This approach enables explanations to follow intrinsic image structure, improving robustness compared to grid-based or pixel-wise attribution methods.
🧪 Experimental Setup
- Task: Explainable Anomaly Detection
- Method: ShapBPT
- Model Type: Black-box anomaly detection models (e.g., VAE-GAN)
- Explanation Level: Pixel-level and region-level attribution
👥 Authors
| Sr. No. | Author Name | Affiliation | Google Scholar |
|---|---|---|---|
| 1. | Muhammad Rashid | University of Torino, Dept. of Computer Science, Torino, Italy | Muhammad Rashid |
| 2. | Elvio G. Amparore | University of Torino, Dept. of Computer Science, Torino, Italy | Elvio G. Amparore |
| 3. | Enrico Ferrari | Rulex Innovation Labs, Rulex Inc., Genova, Italy | Enrico Ferrari |
| 4. | Damiano Verda | Rulex Innovation Labs, Rulex Inc., Genova, Italy | Damiano Verda |
🖼️ Sample Output
🔑 Keywords
ShapBPT · Explainable Anomaly Detection · XAI · Shapley Values · Binary Partition Trees · Computer Vision · ICPE 2026
📖 Citation (BibTeX)
@inproceedings{rashid2026shapbptperspective,
title = {ShapBPT in Perspective: A Consolidated Review and an eXplainable Anomaly Detection Case Study},
author = {Rashid, Muhammad and Amparore, Elvio G.},
booktitle = {Proceedings of the QualITA Workshop at ICPE 2026},
year = {2026},
publisher = {ACM},
doi = {10.1145/3777911.3800638},
url = {https://doi.org/10.1145/3777911.3800638}
}