El grupo de Tecnologías Fluidodinámicas (TFD) del I3A organiza los próximos días 2 y 9 de septiembre el curso "On Two New Ideas for Nonlinear Model Reduction".
Os hacemos llegar información detallada del mismo por si es de vuestro interés.
ON TWO NEW IDEAS FOR NONLINEAR MODEL REDUCTION
Donsub Rim
Assistant Professor of Mathematics,
Department of Mathematics and Statistics,
Washington University in St. Louis.
(https://math.wustl.edu/people/donsub-rim)
DESCRIPTION:
This short course, funded by the Escuela de Doctorado de la Universidad de Zaragoza, is intended to give an informal introduction to two recent ideas introduced to generalize linear reduced models. These ideas were devised to overcome the fundamental limitations of linear reduced models when applied to wave phenomena, and they necessarily involve nonlinear aspects. One is that of reduced deep networks (RDNs), a feedforward deep neural network with a low-rank structure in the weights and biases. The other is the Radon transform to extend these ideas to multiple spatial dimensions. In particular, we will discuss the approximate discrete Radon transform (ADRT) which is a discrete multi-scale approximation.
The lectures will be organized into two parts:
- 02/09/2021 17:00-19:00: Part I, will provide a mathematical description of the challenges the linear reduced models face, in the form of lower bounds for the Kolmogorov N-width. Then simplest nonlinear reduced models will be introduced. (meet.google.com/tqy-dzqh-ste).
- 07/09/2021 17:00-19:00: Part II, will define the RDN and show that this definition provides a flexible framework with which one can describe nonlinear reduced models, and further discuss how the ADRT can be used in conjunction with the RDNs to achieve model reduction for multi-dimensional wave phenomena. (meet.google.com/wpr-xwsz-mqf).