Machine learning helps researchers gain crucial understanding into quantum foundations

Machine learning is a versatile tool useful for solving a wide range of scientific and engineering problems. Bharti et al. present a brief review on how machine learning applications in sequential decision making, classification and regression might be applied to quantum foundations.

The review begins by detailing the basics of machine learning and quantum foundations. It then describes recent research on the intersection of the two subjects.

“Machine learning could be used to discover noise-robust Bell inequalities, find near-optimum configurations to harness quantum resources and understand the geometry of numerous complicated sets which arise in quantum foundations,” said author Tobias Haug.

The authors focused on quantum theory problems such as the determination of the quantum bound for Bell inequalities, training AI for playing Bell nonlocal games, the classification of different behaviors in local/nonlocal sets, and using hidden neurons as hidden variables for completion of quantum theory.

The final part of the survey discusses potential applications of machine learning in quantum foundations, for example, using it to help solve the difficult challenge of designing optimal Bell inequalities for many-body systems. The authors also addressed the possibility of using machine learning in a black-box approach to quantum theory as a pathway to study the classic Turing Test.

“Machine learning could be used in complicated non-convex scenarios as those emerging in a quantum internet. Techniques from multi-agent reinforcement learning could be useful in such cases,” said author Kishor Bharti. “I am hopeful that as complexity increases, machine learning will turn out to be a handy tool.”

Source: “Machine learning meets quantum foundations: A brief survey,” by Kishor Bharti, Tobias Haug, Vlatko Vedral, and Leong-Chuan Kwek, AVS Quantum Science (2020). The article can be accessed at https://doi.org/10.1116/5.0007529 .