Research map / Learning theory and algorithms / Prediction and decision losses
Tilted Sparse Additive Models
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 negative tilt (t<0) instead of ERM-like small tilt in T-SpAM(F) | T-SpAM(F) regression · Synthetic regression, noise ϵB | ASE: 0.350 (SpAM) to 0.019 (t=-2.0) | Helped |
| Tilted loss (TERM) with random Fourier features vs ERM sparse additive model | T-SpAM(F) regression · Synthetic regression, noise ϵA | ASE: 0.204 to 0.173 | Helped |
| Random Fourier feature acceleration vs kernel matrix computation (ablation) | T-SpAM · Synthetic regression, noise ϵA and ϵB | runtime and ASE | No clear effect |
| Tilted logistic loss (T-SpAM) vs ERM SpAM on contaminated labels | T-SpAM(F) classification · Robust classification, r1=0.1 | accuracy: 0.791 to 0.887 | Helped |
| Tilted logistic loss vs ERM SpAM on CME real data | T-SpAM(F) regression · CME arrival-time data | AMSE: 1.037 to 0.615 | Helped |
| Large positive tilt (t=0.5 to 2.0) vs negative tilt on noise ϵA regression (ablation) | T-SpAM(F) · Synthetic regression, noise ϵA | ASE: 0.173 (t=0.1) rising to 0.296 (t=2.0) | Hurt |
| Stronger negative tilt on heavy-tailed noise ϵC (ablation) | T-SpAM(F) · Synthetic regression, noise ϵC | ASE: 0.041 (t=-0.1) vs 0.084 (t=-2.0) | Mixed |
Closest papers by meaning
- Tilted Empirical Risk Minimization (ICLR 2021 · similarity 0.87)
- Generalization and Robustness of the Tilted Empirical Risk (ICML 2025 · similarity 0.87)
- Sparse Shrunk Additive Models (ICML 2020 · similarity 0.86)
- Wavelet regression and additive models for irregularly spaced data (NeurIPS 2018 · similarity 0.85)
- Robust Sparse Regression with Non-Isotropic Designs (NeurIPS 2024 · similarity 0.83)
- Bayesian Nonparametrics Meets Data-Driven Distributionally Robust Optimization (NeurIPS 2024 · 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.