CCSS Meeting #60: Neural Differential Equations and their Application to Atmosphere Models

This lecture will be held in hybrid format: the speaker Dr. Maximilian Gelbrecht will be speaking online from Germany, and Prof. dr. Henk Dijkstra will be physically present as the on-site moderator to open the session and initiate discussions. The theme of this CCSS Lunch Meeting is Neural Differential Equations and the applications in Complex Systems.

In the CCSS living room, participants can enjoy refreshments before and dinner afterwards - please signup for free below.

Speaker Overview

Dr. Maximilian Gelbrecht is a researcher at TU Munich and the FutureLab Artificial Intelligence of the Potsdam Institute for Climate Impact Research (PIK). He previously completed my PhD at Humboldt University of Berlin, studying Network Science, Dynamical Systems and Machine Learning. He is currently working on the application of Neural Partial Differential Equations to Atmospheric Dynamics.

Lecture Overview

When predicting complex systems such as parts of the Earth system, one typically relies on differential equations which often can be incomplete, missing unknown influences or include errors through their discretization. Using the universal differential equations or neural differential equation framework, we can augment the equations with artificial neural networks (ANNs) that can compensate these deficiencies by learning them from data. In the talk, I’ll show that this can be used to predict the dynamics of high-dimensional spatiotemporally chaotic partial differential equations, such as the ones describing atmospheric dynamics. In particular, I will talk about ongoing work that set ups a hybrid model that incorporates both a knowledge-based part of an atmosphere model and a data-driven part, an ANN. Its process-based parts stem from the well studied Marshall Molteni Quasigeostrophic Model. I will talk about the capabilities of this hybrid model and give an outlook of the potential of hybrid models and NDEs for Earth system modelling in general.

Meeting Details

There will be 45-min lecture from the speaker, followed by a 15-min Question & Answer session.

To attend the lecture online, please click the Zoom link at 12:00 on Tuesday 7 November.

To attend the lecture & lunch (physically), please signup below before 12:00 on Monday 6 November.