Research map / Learning theory and algorithms / Matrix and tensor methods
Understanding Trainable Sparse Coding with Matrix Factorization
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 learned LISTA layers instead of standard ISTA iterations | Sparse coding on Gaussian dictionaries · Synthetic Bernoulli-Gaussian sparse coding | sparse-coding objective gap | Helped |
| Use FacNet's constrained matrix-factorization layers instead of standard ISTA iterations | Sparse coding on Gaussian dictionaries · Synthetic Bernoulli-Gaussian sparse coding | sparse-coding objective gap | Helped |
| Use LISTA instead of FacNet | Sparse coding on Gaussian dictionaries · Synthetic Bernoulli-Gaussian sparse coding | sparse-coding objective gap | Helped |
| Add more learned layers (ablation) | LISTA, LFISTA, and FacNet · Synthetic sparse coding | sparse-coding objective gap | Mixed |
| Use adaptive LISTA or FacNet layers beyond the initial starting-point update | Sparse coding with an adversarial Fourier dictionary · Synthetic adversarial dictionary | sparse-coding objective gap | No clear effect |
| Use LISTA or FacNet instead of standard ISTA | Sparse coding with a translation-invariant Haar-wavelet dictionary · PASCAL VOC 2008 image patches | sparse-coding objective gap | Helped |
| Use LISTA or FacNet instead of standard ISTA | Sparse coding with a learned dictionary · MNIST | sparse-coding objective gap | Helped |
| Use a learned linear encoder only as the initialization for ISTA, rather than adaptive learned layers (ablation) | Sparse coding on Gaussian dictionaries · Synthetic Bernoulli-Gaussian sparse coding | sparse-coding objective gap | Mixed |
| Use a denser coefficient distribution (ablation) | Learned sparse coding networks · Synthetic sparse coding | approximation precision | Helped |
| Use a denser coefficient distribution (ablation) | Learned sparse coding networks · Synthetic sparse coding | initial acceleration | Hurt |
| Use a more constrained input distribution (ablation) | LISTA and FacNet · MNIST compared with natural-image patches | sparse-coding objective gap | Helped |
Closest papers by meaning
- Sparse Coding with Gated Learned ISTA (ICLR 2020 · similarity 0.89)
- Provable Acceleration of Nesterov's Accelerated Gradient for Asymmetric Matrix Factorization and Linear Neural Networks (NeurIPS 2024 · similarity 0.88)
- Two Sparse Matrices are Better than One: Sparsifying Neural Networks with Double Sparse Factorization (ICLR 2025 · similarity 0.87)
- Deep Weight Factorization: Sparse Learning Through the Lens of Artificial Symmetries (ICLR 2025 · similarity 0.86)
- On the Crucial Role of Initialization for Matrix Factorization (ICLR 2025 · similarity 0.85)
- Navigating Extremes: Dynamic Sparsity in Large Output Spaces (NeurIPS 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.