Research map / Learning theory and algorithms / Kernels and regression theory
Binarsity: a penalization for one-hot encoded features in linear supervised learning
Mokhtar Z. Alaya, Simon Bussy, Stéphane Gaïffas, Agathe Guilloux
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 |
|---|---|---|---|
| add total-variation penalty within groups (binarsity instead of Lasso / Group L1) (ablation) | logistic regression on binarized features · 9 UCI datasets | AUC | Improved |
| add the extra linear constraint (binarsity instead of Group TV) (ablation) | group TV penalization · 9 UCI datasets | AUC and training time: Group TV AUC always slightly below binarsity, much larger training time | Improved |
| increase number of bins d_j beyond 50 (ablation) | binarsity · Adult and Default of credit card | AUC and training time: performance roughly constant past 50 bins, training time increases | No clear effect |
| use binarsity instead of GAM | linear classifier · 9 UCI datasets | AUC | Improved |
| use binarsity instead of random forest or gradient boosting | classifier · HIGGS | training time: more than 100 times faster | Improved |
| use binarsity instead of l1 penalization | logistic regression · 9 UCI datasets | training time: between 2 and 5 times slower | Hurt |
Closest papers by meaning
- Learning Sparse Classifiers: Continuous and Mixed Integer Optimization Perspectives (JMLR 2021 · similarity 0.85)
- Adaptive Regularization for Large-Scale Sparse Feature Embedding Models (ICLR 2026 · similarity 0.85)
- Solving L1-regularized SVMs and Related Linear Programs: Revisiting the Effectiveness of Column and Constraint Generation (JMLR 2022 · similarity 0.85)
- Recovery of sparse linear classifiers from mixture of responses (NeurIPS 2020 · similarity 0.85)
- Enhanced Feature Learning via Regularisation: Integrating Neural Networks and Kernel Methods (JMLR 2025 · similarity 0.84)
- Global Minimizers of $\ell^p$-Regularized Objectives Yield the Sparsest ReLU Neural Networks (NeurIPS 2025 · 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.
