About the project
Recognising patterns in data to simulate their distributions is a machine learning task that can be enhanced by generating measurement samples from quantum states that are suitably prepared by parameterised quantum circuits. By exploiting symmetry properties, the project will build efficient quantum generative models that have a wide range of applications.
The intersection between the fields of machine learning and quantum computing is an exciting arena for the development of novel algorithms. Measurements on quantum states reveal encoded correlations that go beyond what classical random variables can directly represent. This property makes them a promising component in generative models for learning probability distributions with sample efficiency not achievable using classical methods. The target quantum states are created by quantum circuits with parameterised gates, and the parameters are set using machine learning methods on a classical computer, making it a hybrid quantum-classical approach.
This project will make use of symmetries to constrain the search space of parameters and the representations of states that they generate, in order to guard against training bottlenecks Applications of these methods lie in estimating average values of quantities that characterise a wide range of problems, from chemistry to combinatorial optimisation. While breakthroughs in this research direction is anticipated in all these areas and more, the focus of this project will, however, be restricted to generative machine learning. You'll be part of a vibrant research group where reading groups and discussion forums provide additional sources of analytical thinking, problem-solving and presentation skills.