Beyond-classical Machine learning and AI for Quantum Physics

A primary challenge in quantum computing (QC) is finding its ideal application, i.e., an essential problem with the largest advantage of quantum over classical computing. To resolve it, I propose to focus on the notoriously complex area of quantum many-body systems.

This project will characterise which quantum many-body problems, in various physics domains, allow for significant quantum advantages even over any future machine learning, data-driven methods.

By exploiting my pioneering research in this area, I will also develop new quantum machine learning (QML) methods to solve them better than classically possible, using a two-stage approach.In the first stage, we will develop the project's theoretical foundations.

My recent works on quantum-over-classical learning advantages provide the starting points for the development of new mathematical machinery which facilitates the proving of quantum advantages in selected many-body settings.

In parallel,building on circuit-decomposition methods I recently developed, we will elucidate the role of quantum phenomena in QML in order to design new QML methods which can be better tuned to quantum many-body settings.

In the second stage, we will identify suitable concrete quantum many-body problems with substantial real-world interest, apply the newly designed high-performing quantum learners, and formally prove learning advantages using the developed theoretical machinery.

The positive results of the project will resolve some of the main open problems in QML and will have a major impact on both QC theory and aspects of foundations and applications of QML.

In our search for the best application, we will consider many-body problems from diverse areas of physics: condensed matter, high-energy, and quantum control.

The project will therefore also establish new bridges between quantum many-body physics, machine learning, and quantum computing.