17
6

Efficiently Computing Sparse Fourier Transforms of qq-ary Functions

Abstract

Fourier transformations of pseudo-Boolean functions are popular tools for analyzing functions of binary sequences. Real-world functions often have structures that manifest in a sparse Fourier transform, and previous works have shown that under the assumption of sparsity the transform can be computed efficiently. But what if we want to compute the Fourier transform of functions defined over a qq-ary alphabet? These types of functions arise naturally in many areas including biology. A typical workaround is to encode the qq-ary sequence in binary, however, this approach is computationally inefficient and fundamentally incompatible with the existing sparse Fourier transform techniques. Herein, we develop a sparse Fourier transform algorithm specifically for qq-ary functions of length nn sequences, dubbed qq-SFT, which provably computes an SS-sparse transform with vanishing error as qnq^n \rightarrow \infty in O(Sn)O(Sn) function evaluations and O(Sn2logq)O(S n^2 \log q) computations, where S=qnδS = q^{n\delta} for some δ<1\delta < 1. Under certain assumptions, we show that for fixed qq, a robust version of qq-SFT has a sample complexity of O(Sn2)O(Sn^2) and a computational complexity of O(Sn3)O(Sn^3) with the same asymptotic guarantees. We present numerical simulations on synthetic and real-world RNA data, demonstrating the scalability of qq-SFT to massively high dimensional qq-ary functions.

View on arXiv
Comments on this paper