Machine Learning for Accelerating Acoustic Simulations

Описание: This is a recording of an Acoustics Lunchtime Research Seminar given by Dr Antonio Stanziola of the Biomedical Ultrasound Group of UCL on the 17th November 2021. It was given jointly to the Acoutics Research Group at the University of Salford and the Computational Acoustics Special Interest Group of the UK Acoustics Network.

Abstract: Many fields of science rely on large and complex simulations to make real-world predictions, and acoustics is no exception. Simulation algorithms, however, are often designed around general laws to keep a wide domain of applicability, often at the expenses of heavy and long computations. Learning methods, on the other hand, concentrate the effort in the training phase and enable fast predictions, possibly holding the key to unlock real-time applications. In this talk, we will look at how machine learning is making its way into computational acoustics, what are the challenges lying ahead, and how it can be used to make fast predictions of ultrasound fields in the brain for therapeutic applications.

Bio: Antonio is a Research Fellow at the Biomedical Ultrasound Group of UCL, where he works on Deep Learning for transcranial acoustic simulations. Before, he obtained a PhD from Imperial College with a thesis on vascular imaging with ultrasound and microbubbles. His research interests are at the interface of machine learning and physical modelling, inverse problems and differentiable programming.