English

Multiplying Matrices Without Multiplying

Machine Learning 2021-08-17 v1 Hardware Architecture Performance Machine Learning

Abstract

Multiplying matrices is among the most fundamental and compute-intensive operations in machine learning. Consequently, there has been significant work on efficiently approximating matrix multiplies. We introduce a learning-based algorithm for this task that greatly outperforms existing methods. Experiments using hundreds of matrices from diverse domains show that it often runs 100×100\times faster than exact matrix products and 10×10\times faster than current approximate methods. In the common case that one matrix is known ahead of time, our method also has the interesting property that it requires zero multiply-adds. These results suggest that a mixture of hashing, averaging, and byte shuffling-the core operations of our method-could be a more promising building block for machine learning than the sparsified, factorized, and/or scalar quantized matrix products that have recently been the focus of substantial research and hardware investment.

Keywords

Cite

@article{arxiv.2106.10860,
  title  = {Multiplying Matrices Without Multiplying},
  author = {Davis Blalock and John Guttag},
  journal= {arXiv preprint arXiv:2106.10860},
  year   = {2021}
}

Comments

To appear at ICML 2021