Research map / Learning theory and algorithms / Matrix and tensor methods
Heavy-Tailed Analogues of the Covariance Matrix for ICA
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 |
|---|---|---|---|
| Apply HTICA with centroid-body scaling before FastICA | FastICA · Speech data (perimeterreg) | Frobenius error | Helped |
| Apply HTICA with centroid-body scaling before FastICA | FastICA · Speech data (perimeterreg) | signal recovery | Helped |
| Add Gaussian damping (ablation) | FastICA and other ICA algorithms with an orthogonal mixing matrix · Synthetic heavy-tailed data | Frobenius error | Helped |
| Add Gaussian damping (ablation) | FastICA with pow3 contrast · Synthetic data with similar component tail exponents | Frobenius error | Hurt |
| Use covariance-based rather than centroid-body orthogonalization (ablation) | HTICA · Synthetic data with equal or similar tail exponents | Frobenius error | Helped |
| Use centroid-body rather than covariance-based orthogonalization (ablation) | HTICA · Synthetic heavy-tailed data, in general | numerical stability | Helped |
| Use covariance-based orthogonalization instead of centroid-body scaling (ablation) | HTICA · Synthetic data with mixed component tail exponents | Frobenius error | Hurt |
| Skip orthogonalization and damping (ablation) | FastICA · Synthetic heavy-tailed data | Frobenius error | Mixed |
| Use HTICA rather than plain FastICA | FastICA · Synthetic heavy-tailed data | Frobenius error | No clear effect |
| Use covariance-based orthogonalization | ICA with heavy-tailed components having finite (1+γ)-moment · Heavy-tailed ICA | orthogonality | Helped |
| Use centroid-body scaling instead of sampling uniformly from the centroid body | HTICA orthogonalization · Heavy-tailed ICA | computational practicality | Helped |
Closest papers by meaning
- On the Identifiability of Nonlinear ICA: Sparsity and Beyond (NeurIPS 2022 · similarity 0.87)
- The Scaling Limit of High-Dimensional Online Independent Component Analysis (NeurIPS 2017 · similarity 0.87)
- Nonparametric Evaluation of Noisy ICA Solutions (NeurIPS 2024 · similarity 0.85)
- Diverse Influence Component Analysis: A Geometric Approach to Nonlinear Mixture Identifiability (NeurIPS 2025 · similarity 0.84)
- On Sparse Canonical Correlation Analysis (NeurIPS 2024 · similarity 0.83)
- Efficient Sparse PCA via Block-Diagonalization (ICLR 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.