Research map / Probabilistic and scientific modeling / Causal discovery
Determining the Relevance of Features for Deep Neural Networks
Christian Reimers, Jakob Runge, Joachim Denzler
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 |
|---|---|---|---|
| test relevance with causal conditional-independence test instead of r2 | pattern relevance · Synthetic box vs cross toy task | identifying the used pattern: r2 0.985 for non-used pattern vs p-value 0.608 (r2 1.000, p-value 0.000 for kernel pattern) | Improved |
| compare ML-GCN to SRN on area relevance | SRN · MS COCO val 2014 (80 classes) | classes where area is relevant: 69 -> 73 | Improved |
| compare ML-GCN to SRN on vertical position relevance (lower is better) | SRN · MS COCO val 2014 (80 classes) | classes where vertical position is relevant: 54 -> 51 | Improved |
| compare ML-GCN to SRN on horizontal position relevance (lower is better) | SRN · MS COCO val 2014 (80 classes) | classes where horizontal position is relevant: 16 -> 10 | Improved |
| random horizontal flipping augmentation (horizontal vs vertical half) | ML-GCN · MS COCO val 2014 (80 classes) | classes where half is relevant: 1 class horizontal vs 25 classes vertical | Improved |
| yellow lores area or color as feature | LSFGC · CUB200 (White-crowned vs White-throated sparrow, 120 images) | p-value: 0.406 and 0.446 | No clear effect |
| white throat markings area or color as feature | LSFGC · CUB200 (White-crowned vs White-throated sparrow, 120 images) | p-value: 0.404 and 0.330 | No clear effect |
| shape symmetry as feature: seborrheic keratosis vs melanoma | SLA · HAM10000 / ISIC 2017 validation | p-value: 0.632 -> 0.022 | Mixed |
Closest papers by meaning
- CXPlain: Causal Explanations for Model Interpretation under Uncertainty (NeurIPS 2019 · similarity 0.86)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.85)
- Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning (ICLR 2025 · similarity 0.85)
- Neural Collapse in Cumulative Link Models for Ordinal Regression: An Analysis with Unconstrained Feature Model (NeurIPS 2025 · similarity 0.84)
- Testing Semantic Importance via Betting (NeurIPS 2024 · similarity 0.84)
- Neural Causal Abstractions (AAAI 2024 · similarity 0.84)
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.
