VQE-generated dataset for machine learning with possible quantum advantage

In the last few years, researchers have proposed artificial setups in which quantum machine learning has an advantage over classical one. In this study, we propose the problem of clustering quantum circuits generated by variational quantum eigensolver (VQE) trained on different classes of Hamiltonians. We expect this problem to be easy for quantum machine learning but not classical one. This is because, with quantum computers, we can easily compute the similarity of output states by running the circuits and estimating their inner products, but not with classical computers. Furthermore, this problem is meaningful for, for example, a cloud service provider who wants to analyze quantum circuits received from users. We generate a dataset using various ansatzes and Hamiltonians on the classical simulator and compare the performance of quantum machine learning approach with existing classical algorithms.

The dataset will be available on GitHub to improve the benchmarking environment for quantum machine learning algorithms.