Research map / Learning theory and algorithms / Sample complexity
Classification of Heavy-tailed Features in High Dimensions: a Superstatistical Approach
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 fat-tailed inverse-Gamma cloud (finite a) instead of Gaussian limit a->infinity, same covariance | ridge-regularised GLM on superstatistical mixture · balanced two-cluster synthetic data, square loss | test error | Helped |
| use fat-tailed clouds instead of Gaussian with logistic loss | ridge-regularised GLM on superstatistical mixture · balanced clusters, logistic loss | test error | Helped |
| increase shape parameter a (more Gaussian) | ridge-regularised GLM on superstatistical mixture · synthetic data, logistic loss lambda to 0 | interpolation threshold alpha | Mixed |
| use unbalanced clusters instead of balanced | ridge-regularised GLM on superstatistical mixture · synthetic data | test error | Helped |
| increase ridge regularisation to infinity | ridge-regularised GLM on superstatistical mixture · balanced clusters, square loss, a=2 | test error | Helped |
| increase ridge regularisation to infinity | ridge-regularised GLM on superstatistical mixture · unbalanced clusters, rho=1/4 | test error | No clear effect |
| replace class-conditional labels with random labels | ridge-regularised GLM on superstatistical mixture · ridge regression, zero regularisation | training loss | No clear effect |
| replace class-conditional labels with random labels | ridge-regularised GLM on superstatistical mixture · ridge regression | training error and test error | Hurt |
Closest papers by meaning
- Learning Gaussian Mixtures with Generalized Linear Models: Precise Asymptotics in High-dimensions (NeurIPS 2021 · similarity 0.88)
- The Breakdown of Gaussian Universality in Classification of High-dimensional Linear Factor Mixtures (ICLR 2025 · similarity 0.88)
- Near-Optimal Streaming Heavy-Tailed Statistical Estimation with Clipped SGD (NeurIPS 2024 · similarity 0.85)
- A Separation in Heavy-Tailed Sampling: Gaussian vs. Stable Oracles for Proximal Samplers (NeurIPS 2024 · similarity 0.84)
- On the Efficiency of ERM in Feature Learning (NeurIPS 2024 · similarity 0.84)
- Wrapped Gaussian on the manifold of Symmetric Positive Definite Matrices (ICML 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.