Generalized tensor transforms and their applications in classical and quantum computing

TL;DR

This paper introduces Generalized Tensor Transforms (GTTs), a flexible framework that generalizes many transforms, offering exponential speedups in classical and quantum computing, with applications in quantum state compression, encoding, and signal processing.

Contribution

The paper presents a novel GTT framework that generalizes existing transforms, providing efficient algorithms with exponential speedups for classical and quantum applications.

Findings

GTT algorithms achieve exponential speedup over classical FFT and FWHT.

Quantum GTT algorithms have logarithmic gate complexity and depth, outperforming QFT.

Numerical results show improved quantum state compression and encoding performance.

Abstract

We introduce a novel framework for Generalized Tensor Transforms (GTTs), constructed through an $n$-fold tensor product of an arbitrary $b \times b$ unitary matrix $W$. This construction generalizes many established transforms, by providing a adaptable set of orthonormal basis functions. Our proposed fast classical algorithm for GTT achieves an exponentially lower complexity of $O(N \log_b N)$ in comparison to a naive classical implementation that has an associated computational cost of $O(N^2)$. For quantum applications, our GTT-based algorithm, implemented in the natural spectral ordering, achieves both gate complexity and circuit depth of $O(\log_b N)$, where $N = b^n$ denotes the length of the input vector. This represents a quadratic improvement over Quantum Fourier Transform (QFT), which requires $O((\log_b N)^2)$ gates and depth for $n$ qudits, and an exponential advantage over classical Fast Fourier Transform (FFT) based and Fast Walsh-Hadamard Transform (FWHT) based methods, which incur a computational cost of $O(N \log_b N)$. We explore diverse applications of GTTs in quantum computing, including quantum state compression and transmission, function encoding and quantum digital signal processing. The proposed framework provides fine-grained control of the transformation through the adjustable parameters of the base matrix $W$. This versatility allows precise shaping of each basis function while preserving their effective Walsh-type structure, thus tailoring basis representations to specific quantum data and computational tasks. Our numerical results demonstrate that GTTs enable improved performance in quantum state compression and function encoding compared to fixed transforms (such as FWHT or FFT), achieving higher fidelities with fewer retained components. We also provided novel classical and quantum digital signal filtering algorithms based on our GTT framework.