Research map / Architectures and efficiency / RNNs and normalization
FFTLasso: Large-Scale LASSO in the Fourier Domain
Check 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 FFT-based lifted dual updates instead of the DL-ADMM linear-system solver | LASSO · Synthetic square dictionaries with m = n > 3000 | runtime: reaching a 13ˆ speedup | Helped |
| Use FFTLasso-GPU instead of DL-ADMM-GPU | LASSO · YaleB face recognition, m = 2^12, n = 10^4 | time to reach 92.14% accuracy: « 230 seconds versus more than 600 seconds | Helped |
| Use less-downsampled face images with the larger dictionary dimension | FFTLasso and DL-ADMM face recognition · YaleB face recognition | recognition accuracy: 92.14% -> 97.86% | Helped |
| Add Nesterov accelerated dual ascent (Q=30) rather than regular dual ascent (Q=1) (ablation) | FFTLasso-GPU · Synthetic dictionaries, m = n = 2^13 | ADMM iterations to convergence: 35% less iterations | Helped |
| Add Nesterov accelerated dual ascent (Q=30) rather than regular dual ascent (Q=1) (ablation) | DL-ADMM-GPU · Synthetic dictionaries, m = n = 2^13 | ADMM iterations to convergence: 13% | Helped |
| Use FFTLasso-GPU with Nesterov acceleration instead of conventional DL-ADMM-GPU | LASSO · Synthetic dictionaries, m = n = 2^13 | runtime: 37% | Helped |
| Increase face-image dimension from m = 2^12 to m = 2^13 | YaleB face recognition · YaleB face recognition, n = 10^4 | time to reach target recognition accuracy: « 230 seconds -> « 240 seconds; more than 600 seconds -> « 240 seconds | Mixed |
Closest papers by meaning
- MPGL: An Efficient Matching Pursuit Method for Generalized LASSO (AAAI 2017 · similarity 0.85)
- SparseFormer: Sparse Visual Recognition via Limited Latent Tokens (ICLR 2024 · similarity 0.85)
- Global Linear and Local Superlinear Convergence of IRLS for Non-Smooth Robust Regression (NeurIPS 2022 · similarity 0.85)
- PAC-FNO: Parallel-Structured All-Component Fourier Neural Operators for Recognizing Low-Quality Images (ICLR 2024 · similarity 0.85)
- On the Learn-to-Optimize Capabilities of Transformers in In-Context Sparse Recovery (ICLR 2025 · similarity 0.83)
- Fourier-basis Functions to Bridge Augmentation Gap: Rethinking Frequency Augmentation in Image Classification (CVPR 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.