Equivariant Machine Learning for Physics Modelling
Symmetries play a paramount role in the mathematical sciences. For example, they describe the crystal structure of atoms in materials, and understanding symmetries often goes a long way towards the solution of a problem. In this talk I will introduce the field of equivariant neural networks which studies how to construct symmetric architectures and discuss an application of these tools to a problem in statistical physics.
Speaker Bio – Dr Roberto Bondesan
The goal of my research is to understand how quantum computing and machine learning can help us solve hard computational problems that occur in science and engineering, in particular combinatorial optimization problems and the simulation of quantum systems.
I have a PhD in theoretical physics from UPMC Paris, and in the early stages of my career I worked on the study of topological phases of quantum matter at the universities of Cologne and Oxford. My work aimed at the design of novel materials for electronics and for quantum computing. I then moved to Qualcomm AI research to work on quantum machine learning and the application of AI to design optimization problems. At Qualcomm, I developed a theory of quantum neural networks for photonic quantum computers, and I led a team of machine learning researchers applying Bayesian optimization, reinforcement learning and graph neural networks to the design of chips and compilers. In 2023 I joined the department of computing at Imperial College London as a senior lecturer in quantum computing. I am also a member of Imperial X.
Some of my ongoing research projects are: machine learning for simulating quantum physical systems; machine learning for quantum error correction; quantum algorithms for combinatorial optimization.
Time: 14.00 – 15.00
Date: Tuesday 24 October
Location: I-X 5 Meeting Room, Level 5
Translation and Innovation Hub (I-HUB)
Imperial White City Campus
84 Wood Lane
Any questions and please contact Eileen Boyce (firstname.lastname@example.org) or Lauren Burton (email@example.com).