Weight Space Correlation Analysis: Quantifying Feature Utilization in Deep Learning Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917247458476032 |
|---|---|
| author | Wong, Chun Kit Pegios, Paraskevas Weng, Nina Sejer, Emilie Pi Fogtmann Tolsgaard, Martin Grønnebæk Christensen, Anders Nymark Feragen, Aasa |
| author_facet | Wong, Chun Kit Pegios, Paraskevas Weng, Nina Sejer, Emilie Pi Fogtmann Tolsgaard, Martin Grønnebæk Christensen, Anders Nymark Feragen, Aasa |
| contents | Deep learning models in medical imaging are susceptible to shortcut learning, relying on confounding metadata (e.g., scanner model) that is often encoded in image embeddings. The crucial question is whether the model actively utilizes this encoded information for its final prediction. We introduce Weight Space Correlation Analysis, an interpretable methodology that quantifies feature utilization by measuring the alignment between the classification heads of a primary clinical task and auxiliary metadata tasks. We first validate our method by successfully detecting artificially induced shortcut learning. We then apply it to probe the feature utilization of an SA-SonoNet model trained for Spontaneous Preterm Birth (sPTB) prediction. Our analysis confirmed that while the embeddings contain substantial metadata, the sPTB classifier's weight vectors were highly correlated with clinically relevant factors (e.g., birth weight) but decoupled from clinically irrelevant acquisition factors (e.g. scanner). Our methodology provides a tool to verify model trustworthiness, demonstrating that, in the absence of induced bias, the clinical model selectively utilizes features related to the genuine clinical signal. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_13144 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Weight Space Correlation Analysis: Quantifying Feature Utilization in Deep Learning Models Wong, Chun Kit Pegios, Paraskevas Weng, Nina Sejer, Emilie Pi Fogtmann Tolsgaard, Martin Grønnebæk Christensen, Anders Nymark Feragen, Aasa Computer Vision and Pattern Recognition Machine Learning Image and Video Processing Deep learning models in medical imaging are susceptible to shortcut learning, relying on confounding metadata (e.g., scanner model) that is often encoded in image embeddings. The crucial question is whether the model actively utilizes this encoded information for its final prediction. We introduce Weight Space Correlation Analysis, an interpretable methodology that quantifies feature utilization by measuring the alignment between the classification heads of a primary clinical task and auxiliary metadata tasks. We first validate our method by successfully detecting artificially induced shortcut learning. We then apply it to probe the feature utilization of an SA-SonoNet model trained for Spontaneous Preterm Birth (sPTB) prediction. Our analysis confirmed that while the embeddings contain substantial metadata, the sPTB classifier's weight vectors were highly correlated with clinically relevant factors (e.g., birth weight) but decoupled from clinically irrelevant acquisition factors (e.g. scanner). Our methodology provides a tool to verify model trustworthiness, demonstrating that, in the absence of induced bias, the clinical model selectively utilizes features related to the genuine clinical signal. |
| title | Weight Space Correlation Analysis: Quantifying Feature Utilization in Deep Learning Models |
| topic | Computer Vision and Pattern Recognition Machine Learning Image and Video Processing |
| url | https://arxiv.org/abs/2512.13144 |