Research map / Graph learning / Graph algorithms and clustering
Differentiable Unsupervised Feature Selection based on a Gated Laplacian
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use DUFS instead of CAE | clustering on GISETTE · GISETTE | clustering accuracy: 77.3 -> 99.5 | Helped |
| Use DUFS instead of CAE | clustering on PIX10 · PIX10 | clustering accuracy: 94.1 -> 88.4 | Hurt |
| Use DUFS instead of CAE | clustering on COIL20 · COIL20 | clustering accuracy: 65.6 -> 65.8 | Helped |
| Use DUFS instead of SRCFS | clustering on Yale · Yale | clustering accuracy: 46.3 -> 47.9 | Helped |
| Use DUFS instead of LS | clustering on TOX-171 · TOX-171 | clustering accuracy: 47.5 -> 49.1 | Helped |
| Use DUFS instead of LLCFS | clustering on ALLAML · ALLAML | clustering accuracy: 77.8 -> 74.5 | Hurt |
| Use DUFS instead of SRCFS | clustering on PROSTATE · PROSTATE | clustering accuracy: 60.6 -> 64.7 | Helped |
| Use DUFS instead of NDFS | clustering on RCV1 · RCV1 | clustering accuracy: 55.1 -> 60.2 | Helped |
| Use DUFS instead of CAE | clustering on ISOLET · ISOLET | clustering accuracy: 63.8 -> 63.6 | Hurt |
| Compute the Laplacian on gated features rather than the full feature set | DUFS in high-noise settings · data with nuisance features | clustering performance | Helped |
| Raise the Laplacian to power t=2 rather than t=1 or t=3 (ablation) | DUFS with parameter-free loss on two-moons · two-moons with nuisance dimensions | clustering accuracy | Helped |
Closest papers by meaning
- Unsupervised Feature Selection with Structured Graph Optimization (AAAI 2016 · similarity 0.87)
- Consensus Guided Unsupervised Feature Selection (AAAI 2016 · similarity 0.87)
- Unsupervised Feature Selection Through Group Discovery (AAAI 2026 · similarity 0.86)
- Nonlinear Laplacians: Tunable principal component analysis under directional prior information (NeurIPS 2025 · similarity 0.83)
- SAND: One-Shot Feature Selection with Additive Noise Distortion (ICML 2025 · similarity 0.83)
- Manifold Random Features (NeurIPS 2026 · similarity 0.83)
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.