Research map / Learning theory and algorithms / Matrix and tensor methods
Hiding Data Helps: On the Benefits of Masking for Sparse Coding
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 |
|---|---|---|---|
| Adding masking (decode on subset of dimensions, compute reconstruction loss on complement) instead of standard reconstruction objective | sparse coding with over-realized dictionary · synthetic Gaussian sparse coding, scaling p' (d=100, p=200, k=5) | dictionary recovery error d_R(A,B) | Helped |
| Increasing over-realization p' from p to n under standard objective | baseline reconstruction objective · synthetic, d=100, p=200, k=5, noise var 1/d | dictionary recovery error d_R(A,B) | Hurt |
| Setting p'=2p (modest over-realization) versus p'=p with sample initialization | masking objective · synthetic, d=100, p=200, k=5 | dictionary recovery error d_R(A,B) | Helped |
| Scaling all parameters (d in 100-250, p=2d, k=d/20, p'=2p) with masking vs baseline | masking objective vs baseline · synthetic scaled d, p, k | dictionary recovery error d_R(A,B) | Helped |
| Lowering noise variance from 1/d to 1/d^2 (p'=1000) (ablation) | baseline vs masking · synthetic, low-noise regime | dictionary recovery error d_R(A,B) | No clear effect |
| Masking objective in low-noise regime vs standard reconstruction (ablation) | sparse coding · synthetic, noise std 0.01 to 0.1 | dictionary recovery error d_R(A,B) | No clear effect |
Closest papers by meaning
- A Provable Approach for Double-Sparse Coding (AAAI 2018 · similarity 0.89)
- Dictionary Learning Based on Sparse Distribution Tomography (ICML 2017 · similarity 0.89)
- On the Sparsity-Storage-Accuracy Tradeoff in Parsimoniously Activated Dictionary Learning (NeurIPS 2026 · similarity 0.85)
- Recovery Guarantees for Posterior Sampling of One-Bit Compressed Sensing (NeurIPS 2026 · similarity 0.85)
- The Price of Sparsity: Sufficient Conditions for Sparse Recovery using Sparse and Sparsified Measurements (NeurIPS 2025 · similarity 0.85)
- Compression of Structured Data with Autoencoders: Provable Benefit of Nonlinearities and Depth (ICML 2024 · similarity 0.85)
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.