Researchers from the University of Bristol have developed an algorithm that leverages machine learning to help characterize quantum systems. Building models of quantum systems has long been computationally challenging. By feeding experiment results to the ML-informed simulation, the researchers were able develop increasingly accurate models of the system. This should prove useful when applied to a variety of quantum systems and also provide insight for using such models on real-world quantum computers and sensing devices.
In their paper, Learning models of quantum systems from experiments, published today in Nature Physics, researchers from Bristol’s Quantum Engineering Technology Lab (QETL) describe their method which used an autonomous agent, using machine learning to reverse engineer Hamiltonian models.
Quick summary of problem and solution provided by the researchers:
“In Physics, systems of particles and their evolution are described by mathematical models. In general, retrieving such a model is a difficult problem, requiring the successful interplay of theoretical arguments and experimental verification. Validating proposed models requires expertise in both domains through a lengthy, iterative process. At the quantum mechanical level, the dynamics of a system of particles interacting with each other is often described by a Hamiltonian model. The process of formulating Hamiltonian models from observations is made even harder by the nature of quantum states, which collapse when we try to inspect them.
“We proposed a new protocol to formulate and validate approximate models for quantum systems of interest. Our algorithm works like an autonomous agent, using machine learning to reverse engineer Hamiltonian models. The agent designs and performs experiments on the targeted quantum system, and the resultant data are fed back into the algorithm. It proposes candidate Hamiltonian models to describe the target system, and distinguishes between them using statistical metrics, namely Bayes factors. We successfully demonstrated the algorithm’s ability on a real-life quantum experiment involving defect centers in diamond, which are a well-studied platform for quantum information processing and quantum sensing.”
“Combining the power of today’s supercomputers with machine learning, we were able to automatically discover structure in quantum systems. As new quantum computers/simulators become available, the algorithm becomes more exciting: first it can help to verify the performance of the device itself, then exploit those devices to understand ever-larger systems.” said Brian Flynn of Bristol’s QETLabs in today’s announcement.
“This level of automation makes it possible to entertain myriads of hypothetical models before selecting an optimal one, a task that would be otherwise daunting for systems whose complexity is ever increasing,” added Andreas Gentile, formerly of QETLabs, now at Qu & Co.
The next step, say the researchers, is to extend the algorithm to explore larger systems, and different classes of quantum models which represent different physical regimes or underlying structures.
Feature Image caption: The nitrogen vacancy center set-up, that was used for the first experimental demonstration of QMLA.