Research map / Label-efficient and robust learning / Recognition and long-tail
Jointly Informative Feature Selection Made Tractable by Gaussian Modeling
Leonidas Lefakis, François Fleuret
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 |
|---|---|---|---|
| use joint-informativeness feature selection (GC.MI) instead of marginal InfoGain | GC.MI with linear SVM · STL-10, 100 features | test accuracy: 37.84 -> 48.88 (GC.MI 48.88 vs InfoGain 37.84) | Improved |
| use GKL.E instead of InfoGain | GKL.E with RBF SVM · CIFAR-10, 100 features | test accuracy: 53.94 -> 65.64 | Improved |
| use GC.MI instead of InfoGain | GC.MI with QDA · INRIA, 100 features | test accuracy: 91.77 -> 96.31 | Improved |
| select second feature by joint information instead of marginal (ablation) | GC-approximation of mutual information · synthetic 5-feature data | feature ranking: 49.2% | Improved |
| use GC.MI | GC.MI · 48 dataset-classifier-size combinations | top-three rank: 24 out of 48 top three, first 8 times | Improved |
| use GKL.E | GKL.E · 48 dataset-classifier-size combinations | top-three rank: 27 out of 48 top three, first 11 times | Improved |
| use GC.MI instead of SBMLR for running time | GC.MI · CIFAR-10 feature selection CPU time | CPU time: 1449 -> 20 | Improved |
| use GC.MI instead of MRMR on INRIA | GC.MI with AdaBoost · INRIA, 10 features | test accuracy: 81.53 -> 95.04 | Improved |
| use GC.MI instead of MRMR for running time on INRIA | GC.MI · INRIA CPU time | CPU time: 32 vs 43 | Hurt |
| use GKL.MI instead of GKL.E (ablation) | GKL.MI · CIFAR-10, AdaBoost, 100 features | test accuracy: 56.41 -> 51.12 | Hurt |
Closest papers by meaning
- Variational Information Maximization for Feature Selection (NeurIPS 2016 · similarity 0.87)
- Estimating Conditional Mutual Information for Dynamic Feature Selection (ICLR 2024 · similarity 0.87)
- A Rigorous Information-Theoretic Definition of Redundancy and Relevancy in Feature Selection Based on (Partial) Information Decomposition (JMLR 2023 · similarity 0.86)
- SAND: One-Shot Feature Selection with Additive Noise Distortion (ICML 2025 · similarity 0.86)
- Semi-Supervised Sparse Gaussian Classification: Provable Benefits of Unlabeled Data (NeurIPS 2024 · similarity 0.85)
- Feature Selection for Latent Factor Models (CVPR 2025 · 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.
