Siddhartha Mishra: Deep learning in Scientific computing
Speaker: Siddhartha Mishra (ETH Zürich, UiO)
The lecture on Wednesday, April 24th has been CANCELLED. The lecture on
Thursday, April 25th will start at 13:15!
Abstract: Machine learning has been extremely successful in recent years at diverse tasks in computer science ranging from image classification and natural language processing to game intelligence and protein folding. Machine learning, particularly deep learning, is being increasingly used in scientific computing in a variety of contexts.
The aim of this set of lectures would be to introduce some key concepts in deep learning and highlight a few selected applications to scientific computing. We will introduce deep neural networks and describe different aspects of their architectures. Some theoretical results that underpin the superb performance of deep neural networks will be presented. We conclude by describing the application of deep learning to the numerical solutions of partial differential equations, arising in mathematical finance and in computational fluid dynamics.