Optimal Quantization for Matrix Multiplication
This work addresses the computational bottleneck of matrix multiplication in machine learning, particularly for large models, by developing an optimal quantization method, though it is incremental as it builds on prior compression algorithms.
The paper tackles the problem of lossy compression for matrix multiplication, a key bottleneck in large language models, by providing a non-asymptotic lower bound on approximation error and constructing a universal quantizer based on nested lattices that achieves this bound for iid Gaussian matrices, with a practical version performing close to optimal and revealing a phase-transition at 0.906 bits/entry.
Recent work in machine learning community proposed multiple methods for performing lossy compression (quantization) of large matrices. This quantization is important for accelerating matrix multiplication (main component of large language models), which is often bottlenecked by the speed of loading these matrices from memory. Unlike classical vector quantization and rate-distortion theory, the goal of these new compression algorithms is to be able to approximate not the matrices themselves, but their matrix product. Specifically, given a pair of real matrices $A,B$ an encoder (compressor) is applied to each of them independently producing descriptions with $R$ bits per entry. These representations subsequently are used by the decoder to estimate matrix product $A^\top B$. In this work, we provide a non-asymptotic lower bound on the mean squared error of this approximation (as a function of rate $R$) for the case of matrices $A,B$ with iid Gaussian entries. Algorithmically, we construct a universal quantizer based on nested lattices with an explicit guarantee of approximation error for any (non-random) pair of matrices $A$, $B$ in terms of only Frobenius norms $\|\bar{A}\|_F, \|\bar{B}\|_F$ and $\|\bar{A}^\top \bar{B}\|_F$, where $\bar{A},\bar{B}$ are versions of $A,B$ with zero-centered columns, respectively. For iid Gaussian matrices our quantizer achieves the lower bound and is, thus, asymptotically optimal. A practical low-complexity version of our quantizer achieves performance quite close to optimal. In addition, we derive rate-distortion function for matrix multiplication of iid Gaussian matrices, which exhibits an interesting phase-transition at $R\approx 0.906$ bit/entry, showing necessity of Johnson-Lindestrauss dimensionality reduction (sketching) in the low-rate regime.