Deep Learning meets PDE Workshop
Overview
Deep learning has revolutionized the landscape of PDE computation. On one hand, many challenging problems arising in the physical sciences are governed by PDEs involving high-dimensional solution spaces or high-dimensional parameter fields, where traditional numerical methods face severe computational bottlenecks. Deep neural networks offer a potentially powerful and efficient new paradigm for tackling such problems. On the other hand, many mathematical models contain unknown parameters or missing terms. Deep learning provides a flexible framework to seamlessly incorporate observational data into model prediction, while offering strategies to quantify and address model uncertainties. Despite these promising developments, the indiscriminate use of deep learning for PDE computation poses significant challenges. Training can be expensive and unstable, and may lead to solutions that lack robustness or physical reliability. Moreover, such approaches often rely on large amounts of high-fidelity data, the generation of which is itself a formidable task.
Given the past decades of intellectual development in classical numerical methods for PDEs, there is immense potential in thoughtfully combining deep learning with established computational techniques. This workshop aims to provide a platform for researchers with backgrounds in both classical numerical analysis and modern scientific machine learning to exchange ideas and technical expertise, and fostering dialogue between the scientific computing and machine learning communities.
Registration
Graduate students: in your application, please provide your research interests and research plan, anticipated year of PhD, and an updated CV. Limited funding is available to support participation. If you would like to request funding to participate, please ensure you complete the Funding Request portion of the application.
Organizer
Logistics
Getting to UMN
- The Minneapolis-Saint Paul International Airport (MSP) is conveniently located nearby Downtown Minneapolis. The airport is connected to the city center and to the UMN campus by public transport, particularly the Metro Blue Line.
- Metro Transit is the Minneapolis / St. Paul public transportation system – it includes bus and train services that run from very early morning until late in the evening.
- Bicycles and scooter rentals are available from Lime, Spin, and Veo.
Child care
Care.com connects visiting families with experienced local caregivers.
Explore the Twin Cities
Looking for things to do while in Minneapolis? Here are some great resources:
This workshop is supported by the U.S. National Science Foundation grant #DMS-1846854.
