We introduce new rounding methods to improve the accuracy of finite precision quantum arithmetic. These quantum rounding methods are applicable when multiple samples are being taken from a quantum program. We show how to use multiple samples to stochastically suppress arithmetic error from rounding. We benchmark these methods on the multiplication of fixed-point numbers stored in quantum registers. We show that the gate counts and depths for multiplying to a target accuracy can be reduced by approximately 2-3X over state of the art methods while using roughly the same number of qubits.
@article{arxiv.2108.05949,
title = {Quantum Rounding},
author = {Rajiv Krishnakumar and William Zeng},
journal= {arXiv preprint arXiv:2108.05949},
year = {2021}
}