Research map / Architectures and efficiency / Quantization
Matrix Compression via Randomized Low Rank and Low Precision Factorization
Rajarshi Saha, Varun Srivastava, Mert Pilanci
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| replace naive quantization with LPLR at 1 bit | embedding classifier · CIFAR-10 embeddings | accuracy (%): 11 -> 92 | Improved |
| replace naive quantization with LPLR at 1 bit | embedding classifier · CIFAR-100 embeddings | accuracy (%): 1 -> 79 | Improved |
| replace naive quantization with LPLR at 1 bit | embedding classifier · IMDB embeddings | accuracy (%): 50 -> 73 | Improved |
| replace naive quantization with LPLR at 1 bit | embedding classifier · Emotion embeddings | accuracy (%): 29 -> 41 | Improved |
| reduce LPLR factor bits from B=32 to B=8 (larger sketch) at B_nq=1 (ablation) | LPLR · Shepp-Logan phantom | relative Frobenius norm error: 0.610 -> 0.340 | Improved |
| LPLR instead of naive quantization at B_nq=1, B=8 | LPLR · Shepp-Logan phantom | relative Frobenius norm error: 0.532 -> 0.340 | Improved |
| LPLR instead of naive quantization at B_nq=1, B=32 | LPLR · Shepp-Logan phantom | relative Frobenius norm error: 0.532 -> 0.610 | Hurt |
| LPLR instead of naive quantization at B_nq=2, B=32 | LPLR · Shepp-Logan phantom | relative Frobenius norm error: 0.312 -> 0.447 | Hurt |
| use SVD-based left factor (LPLR-SVD) instead of random sketch (ablation) | LPLR · Shepp-Logan phantom, B=32, m=15 | relative Frobenius norm error: 0.610 -> 0.506 | Improved |
| use SVD-based left factor (LPLR-SVD) instead of random sketch (ablation) | LPLR · IMDB embeddings, B_nq=1 | relative Frobenius norm error: 0.313 -> 0.241 | Improved |
| LPLR instead of direct-SVD quantization | LPLR · CIFAR-100 embeddings, B_nq=1 | accuracy (%): 82 -> 79 | Hurt |
| LPLR instead of direct-SVD quantization | LPLR · IMDB embeddings, B_nq=1 | relative Frobenius norm error: 0.229 -> 0.313 | Hurt |
Closest papers by meaning
- Learning-Based Low-Rank Approximations (NeurIPS 2019 · similarity 0.88)
- Compressing Large Language Models using Low Rank and Low Precision Decomposition (NeurIPS 2024 · similarity 0.86)
- Low Rank Matrix Completion via Robust Alternating Minimization in Nearly Linear Time (ICLR 2024 · similarity 0.86)
- Binary Quadratic Quantization: Beyond First-Order Quantization for Real-Valued Matrix Compression (NeurIPS 2025 · similarity 0.86)
- Fast exact recovery of noisy matrix from few entries: the infinity norm approach (NeurIPS 2025 · similarity 0.85)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.84)
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.
