Research map / Learning theory and algorithms / Matrix and tensor methods
Multiplying Matrices Without Multiplying
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 hash-based encoding and lookup aggregation instead of multiply-add-based approximate products | Approximate matrix multiplication · Approximate matrix products | speedup vs. exact multiplication: 100× | Helped |
| Use Maddness instead of current approximate methods | Approximate matrix multiplication · Approximate matrix products | speedup: 10× | Helped |
| Use Maddness for softmax classifier inference | CIFAR-10 and CIFAR-100 softmax classifiers · CIFAR-10 and CIFAR-100 | classification accuracy at a given runtime | Helped |
| Use Maddness for kernel-based classification | Kernel classifiers · UCR Time Series Archive | classification accuracy at a given runtime | Helped |
| Use Maddness for small image-filter matrix products | Image filtering · Caltech101 | speed-accuracy tradeoff | Helped |
| Apply the Maddness approximation to high-pass rather than low-pass filters | Image filtering · Caltech101 Sobel filters | NMSE | Hurt |
| Optimize prototypes using ridge regression (ablation) | Maddness · CIFAR-10 and CIFAR-100 softmax classifiers | squared error | Helped |
| Optimize prototypes using ridge regression (ablation) | Maddness · UCR Time Series Archive | classification accuracy | Mixed |
| Evaluate more than 4 candidate split indices per tree level (ablation) | MaddnessHash · Hash-function learning | split quality | No clear effect |
| Evaluate only 1 candidate split index per tree level (ablation) | MaddnessHash · Hash-function learning | split quality | No clear effect |
| Use averaging instructions rather than upcasting and adding lookup values | Maddness aggregation · Approximate matrix product aggregation | aggregation speed | Helped |
| Use averaging instructions rather than exact integer summation | Maddness aggregation · Approximate matrix products | accuracy | Hurt |
Closest papers by meaning
- Approximate Multiplication of Sparse Matrices with Limited Space (AAAI 2021 · similarity 0.87)
- LibAMM: Empirical Insights into Approximate Computing for Accelerating Matrix Multiplication (NeurIPS 2024 · similarity 0.87)
- The Structural Complexity of Matrix-Vector Multiplication (NeurIPS 2025 · similarity 0.86)
- IM-Unpack: Training and Inference with Arbitrarily Low Precision Integers (ICML 2024 · similarity 0.84)
- Improved Algorithms for Kernel Matrix-Vector Multiplication Under Sparsity Assumptions (ICLR 2025 · similarity 0.84)
- NestQuant: nested lattice quantization for matrix products and LLMs (ICML 2025 · 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.