Variational quantum algorithms for many-body simulation and machine learning problems

Near-term quantum simulators suffer from various imperfections. A key question is whether such noisy quantum devices can outperform classical computers. Several demonstrations for quantum advantage have been achieved for sampling problems in superconducting and optical platforms.

While these proof of principle experiments show the superiority of quantum computers, they do not offer an immediate practical advantage due to the limited practicality of sampling problems. Variational quantum algorithms are the most promising approach for achieving practical quantum advantage.

These algorithms benefit from a hybrid combination of quantum devices and classical optimizers. In this seminar, we show two distinct applications for such algorithms, namely: (i) quantum simulation of many-body systems; and (ii) machine learning problems.

In the former, we show how symmetries can be harnessed in optimizing circuit design [1] and be implemented experimentally in superconducting quantum simulators [2]. For the latter, a novel error-mitigation algorithm is presented which significantly enhances the performance of variational quantum algorithms for supervised machine learning problems [3].