Quantum computers hold the potential to solve challenging problems in fields such as chemistry and physics, but the lack of quantum error correction currently limits the types of quantum algorithms that can be implemented. To circumvent this limitation, a hybrid approach has been introduced, where classical computers are used in tandem with error-prone quantum computers. Our research extends this hybrid approach by combining quantum computing with classical machine learning (ML) (Figure a), targeting problems that are expected to be difficult to solve using classical computers alone.

We apply this hybrid approach to two important problems in many-body physics. The first is predicting the properties of a system's ground state, which in many cases determines the system's characteristics (Figure b), and the second is identifying various quantum phases, which are distinguished by quantum effects (Figure c).

By applying and developing various error mitigation techniques—which, unlike error correction, can only offset certain types of errors—we were able to obtain refined data from quantum computers. Using this data, we experimentally validated the application of the hybrid approach, which had previously been proposed only theoretically. Through this research, we demonstrate the scalability and effectiveness of hybrid algorithms that utilize both quantum computers and classical computers.

 

Authors: Gyungmin Cho, Dohun Km

Link: https://www.nature.com/articles/s41467-024-51932-3