Research map / Label-efficient and robust learning / Recognition and long-tail
Linking in Style: Understanding learned features in deep learning models
Maren Wehrheim, Pamela Osuna Vargas, Matthias Kaschube
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| link ResNet-50 penultimate layer to StyleGAN-XL W-space with a linear regression linking network | Linking network · ImageNet classes, ResNet-50 and StyleGAN-XL | MSE between w and cycled w, MoCov2 image distance | Improved |
| cluster W-space versus representation space by class with k-means | StyleGAN-XL W-space and ResNet-50 representation space · five random ImageNet classes, 100 repetitions | Adjusted Rand Index | No clear effect |
| use counterfactual quantification pipeline instead of MSE or LPIPS along the trajectory | Decision-boundary analysis · Counterfactual trajectories between dog classes | interpretability of feature changes | Improved |
| apply the linking approach to a ViT classifier instead of a CNN | Linking network · ViT classifiers | applicability | Hurt |
| train linking network on a classifier and GAN trained on similar but different datasets | Linking network · face datasets | ability to reveal relevant features | No clear effect |
Closest papers by meaning
- Network Dissection: Quantifying Interpretability of Deep Visual Representations (CVPR 2017 · similarity 0.90)
- Interpreting Deep Visual Representations via Network Dissection (TPAMI 2018 · similarity 0.89)
- Explaining Deep Convolutional Neural Networks via Latent Visual-Semantic Filter Attention (CVPR 2022 · similarity 0.89)
- GAN Dissection: Visualizing and Understanding Generative Adversarial Networks (ICLR 2019 · similarity 0.88)
- Explaining in Diffusion: Explaining a Classifier with Diffusion Semantics (CVPR 2025 · similarity 0.87)
- Towards flexible perception with visual memory (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
