Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only! Grab it Explore the groundbreaking field of PDE learning in this 59-minute talk by Nicolas Boulle from the University of Cambridge. Delve into the intersection of machine learning, physics, and mathematics as Boulle discusses the data-efficient nature of learning solution operators for elliptic PDEs. Discover how deep learning techniques are utilized to map source terms to PDE solutions, enabling the creation of surrogate data for various applications in engineering and biology. Gain insights into the theoretical explanation behind the surprising data efficiency of PDE learning, including an algorithm that achieves exponential convergence rates. Learn about the potential practical benefits of this research, such as improved dataset and neural network architecture designs. Presented as part of the DDPS (Data-Driven Physical Simulation) series at Lawrence Livermore National Laboratory, this talk offers valuable knowledge for researchers and practitioners in numerical analysis, deep learning, and physical modeling. Read more