Learning fast algorithms for linear transforms using butterfly factorizations

T Dao, A Gu, M Eichhorn, A Rudra… - … conference on machine …, 2019 - proceedings.mlr.press
International conference on machine learning, 2019proceedings.mlr.press
Fast linear transforms are ubiquitous in machine learning, including the discrete Fourier
transform, discrete cosine transform, and other structured transformations such as
convolutions. All of these transforms can be represented by dense matrix-vector
multiplication, yet each has a specialized and highly efficient (subquadratic) algorithm. We
ask to what extent hand-crafting these algorithms and implementations is necessary, what
structural prior they encode, and how much knowledge is required to automatically learn a …
Abstract
Fast linear transforms are ubiquitous in machine learning, including the discrete Fourier transform, discrete cosine transform, and other structured transformations such as convolutions. All of these transforms can be represented by dense matrix-vector multiplication, yet each has a specialized and highly efficient (subquadratic) algorithm. We ask to what extent hand-crafting these algorithms and implementations is necessary, what structural prior they encode, and how much knowledge is required to automatically learn a fast algorithm for a provided structured transform. Motivated by a characterization of fast matrix-vector multiplication as products of sparse matrices, we introduce a parameterization of divide-and-conquer methods that is capable of representing a large class of transforms. This generic formulation can automatically learn an efficient algorithm for many important transforms; for example, it recovers the Cooley-Tukey FFT algorithm to machine precision, for dimensions up to . Furthermore, our method can be incorporated as a lightweight replacement of generic matrices in machine learning pipelines to learn efficient and compressible transformations. On a standard task of compressing a single hidden-layer network, our method exceeds the classification accuracy of unconstrained matrices on CIFAR-10 by 3.9 points—the first time a structured approach has done so—with 4X faster inference speed and 40X fewer parameters.
proceedings.mlr.press
Showing the best result for this search. See all results