Research map / Architectures and efficiency / Quantization
Scalar Quantization as Sparse Least Square Optimization
Chen Wang, Xiaomei Yang, Shaomin Fei, Kai Zhou, Xiao‐Feng Gong, Miao Du, Ruisen Luo
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 |
|---|---|---|---|
| add least-square refit of values to l1 quantization (ablation) | l1 sparse quantization · Last-layer MNIST network weights | l2 information loss / accuracy | Improved |
| add least square refit to l1 quantization (ablation) | l1 sparse quantization · MNIST image quantization | l2 norm loss | Improved |
| add least square refit to l1 quantization (ablation) | l1 sparse quantization · Artificial 1-D data (Gaussian mixture, uniform, single Gaussian) | l2 loss | Improved |
| add negative l2 term to l1 quantization (ablation) | l1 sparse quantization · Last-layer MNIST network weights | number of quantization values | Improved |
| add negative l2 term to l1 quantization (ablation) | l1 sparse quantization · Last-layer MNIST network weights | stability | Hurt |
| replace k-means with clustering-based least square (ablation) | k-means quantization · Last-layer MNIST network weights | accuracy | Improved |
| replace k-means with l1 quantization with least square (ablation) | k-means quantization · Medium-size data | running time | Improved |
| use l1 quantization instead of k-means (ablation) | k-means quantization · Neural network, MNIST, artificial data | l2 information loss | Hurt |
| use l0 optimization quantization (ablation) | l0 sparse quantization · Network weights and artificial data | solution success | Hurt |
| use l1 quantization instead of Mixture of Gaussian quantization (ablation) | Mixture of Gaussian quantization · Last-layer MNIST network weights | accuracy | Improved |
Closest papers by meaning
- Sparse Quantized Spectral Clustering (ICLR 2021 · similarity 0.86)
- Towards the Limit of Network Quantization (ICLR 2017 · similarity 0.86)
- Generalization Error Analysis of Quantized Compressive Learning (NeurIPS 2019 · similarity 0.85)
- K-means clustering using random matrix sparsification (ICML 2018 · similarity 0.84)
- QBB: Quantization with Binary Bases for LLMs (NeurIPS 2024 · similarity 0.84)
- Optimal and Approximate Adaptive Stochastic Quantization (NeurIPS 2024 · similarity 0.82)
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.
