Research map / Architectures and efficiency / Pruning and efficient tuning
Hyperparameter Tuning is All You Need for LISTA
Xiaohan Chen, Jialin Liu, Zhangyang Wang, Wotao Yin
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add momentum to ALISTA (ALISTA-MM) (ablation) | ALISTA · synthetic sparse recovery (m=250,n=500) | NMSE (dB) | Improved |
| use symmetric matrix parameterization (ablation) | ALISTA · synthetic sparse recovery (m=250,n=500) | NMSE (dB) | No clear effect |
| grid search instead of backpropagation for c1,c2 (HyperLISTA-GS vs BP) (ablation) | HyperLISTA · synthetic sparse recovery | NMSE (dB) | Improved |
| also grid-search c3 (HyperLISTA-Full vs GS) (ablation) | HyperLISTA · synthetic sparse recovery | NMSE (dB) | Improved |
| HyperLISTA vs ALISTA, NA-ALISTA, Ada-LFISTA under train/test mismatch (sparsity, magnitude, noise) | HyperLISTA · mismatched synthetic settings | NMSE (dB) | Improved |
| unroll to 40 layers vs ALISTA-Extra (reused last layer) | HyperLISTA · synthetic, 40 layers | NMSE (dB) | Improved |
| HyperLISTA vs ALISTA on natural images | HyperLISTA · Set 11 compressive sensing | PSNR (dB): 32.24 -> 33.46 | Improved |
| HyperLISTA vs FISTA 1,000 iters | HyperLISTA · Set 11 compressive sensing | PSNR (dB): 30.92 -> 33.46 | Improved |
Closest papers by meaning
- Learned Extragradient ISTA with Interpretable Residual Structures for Sparse Coding (AAAI 2021 · similarity 0.87)
- Theoretical Linear Convergence of Unfolded ISTA and Its Practical Weights and Thresholds (NeurIPS 2018 · similarity 0.86)
- Sparse maximal update parameterization: A holistic approach to sparse training dynamics (NeurIPS 2024 · similarity 0.86)
- On the Provable Separation of Scales in Maximal Update Parameterization (ICML 2025 · similarity 0.84)
- SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity (ICML 2025 · similarity 0.83)
- Principled Architecture-aware Scaling of Hyperparameters (ICLR 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.
