Explore an in-depth technical talk by Dr. Marius Zeinhofer from University Hospital Freiburg that delves into infinite-dimensional optimization challenges in scientific machine learning. Learn about the "first optimize, then discretize" paradigm and its application in solving complex optimization problems through neural network frameworks. Discover how state-of-the-art algorithms for scientific machine learning applications are derived and understand the implementation of Kronecker-factored approximation for natural gradient descent in scientific applications. Drawing from his extensive background in multiphysics models and scientific machine learning research, Dr. Zeinhofer presents advanced concepts in error analysis and optimization of neural network-based methods for solving partial differential equations (PDEs), particularly relevant for healthcare AI applications.
Overview
Syllabus
DDPS | “Infinite Dimensional Optimization for Scientific Machine Learning”
Taught by
Inside Livermore Lab
Related Courses
-
Distributed Optimization and Machine Learning
-
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization
5.0 -
Introduction to Machine Learning
3.5 -
Fundamentals of Machine Learning
4.7 -
Machine Learning and NLP Basics
-
Scientific Machine Learning through the Lens of Physics-Informed Neural Networks
