Making the cut: two methods for breaking down a quantum algorithm

Abstract

Despite the promise that fault-tolerant quantum computers can efficiently solve classically intractable problems, it remains a major challenge to find quantum algorithms that may reach computational advantage in the present era of noisy, small-scale quantum hardware. Thus, there is substantial ongoing effort to create new quantum algorithms (or adapt existing ones) to accommodate depth and space restrictions. By adopting a hybrid query perspective, we identify and characterize two methods of ``breaking down'' quantum algorithms into rounds of lower (query) depth, designating these approaches as ``parallelization'' and ``interpolation''. To the best of our knowledge, these had not been explicitly identified and compared side-by-side, although one can find instances of them in the literature. We apply them to two problems with known quantum speedup: calculating the $k$-threshold function and computing a NAND tree. We show that for the first problem parallelization offers the best performance, while for the second interpolation is the better choice. This illustrates that no approach is strictly better than the other, and so that there is more than one good way to break down a quantum algorithm into a hybrid quantum-classical algorithm.