Post-selection in noisy Gaussian boson sampling: part is better than whole

Abstract

Gaussian boson sampling (GBS) is originally proposed to show quantum advantage with quantum linear optical elements. Recently, several experimental breakthroughs based on GBS pointing to quantum computing supremacy have been presented. However, due to technical limitations, the outcomes of GBS devices are influenced severely by photon loss. Here, we present a practical method to reduce the negative effect caused by photon loss. We first show with explicit formulas that a GBS process can be mapped to another GBS processes. Based on this result, we propose a post-selection method which discards low-quality data according to our criterion to improve the performance of the final computational results, say part is better than whole. As an example, we show that the post-selection method can turn a GBS experiment that would otherwise fail in a ‘non-classicality test’ into one that can pass that test. Besides improving the robustness of computation results of current GBS devices, this post-selection method may also benefit the further development of GBS-based quantum algorithms.