Model-driven Machine Learning for Climate and Earth Science
Understanding and predicting the complex dynamics of the Earth’s atmosphere, ocean, land and ice.
This Helmholtz AI young investigator group, led by David Greenberg, is located within the Institute of Coastal Systems - Analysis and Modeling, at the Helmholtz Zentrum Hereon (Hereon). Their research focuses on the Earth system at the boundary between physics-based and machine learning approaches, to gain insights unavailable to either approach alone.
The research group also organizes a weekly seminar on machine learning at Hereon, aimed at the scientific communities of the Helmholtz Institutes and the Hamburg area.
'Model-driven machine learning' group website
Physical models. Numerical simulations based on known physics simulators handle complex systems well, but struggle with data assimilation, parameter tuning and uncertainty quantification.
Machine Learning. Conversely, machine learning techniques can absorb and process large datasets, but typically ignore physics and generalize poorly to new scenarios.
Model-driven Machine Learning. We develop hybrid methods that combine the advantages of deep learning and physical modeling in a Bayesian framework.
- Ramesh, P., Lueckmann, J.-M., Boelts, J., Tejero-Cantero, A., Greenberg, D.S., Gonçalves, P.J., & Macke, J.H. (2022): GATSBI: Generative Adversarial Training for Simulation-Based Inference. International Conference on Learning Representations.
- Nonnenmacher, M., & Greenberg, D.S. (2021): Deep emulators for differentiation, forecasting, and parametrization in Earth science simulators. Journal of Advances in Modeling Earth Systems, 13, e2021MS002554, doi:10.1029/2021MS002554
- Tejero-Cantero, A.; Boelts, J.; Deistler, M.; Lueckmann, J.; Durkan, C.; Goncalves, P.; Greenberg, D.; Macke, J.: sbi: A toolkit for simulation-based inference. In: The Journal of Open Source Software. Vol. 5 (2020) 52, 2505, DOI: /10.21105/joss.0250
- Gonçalves, P. J., Lueckmann, J. M., Deistler, M., Nonnenmacher, M., Öcal, K., Bassetto, G., Chintaluri, C., Podlaski, W. F., Haddad, S. A., Vogels, T. P., Greenberg, D. S., & Macke, J. H. (2020). Training deep neural density estimators to identify mechanistic models of neural dynamics. eLife, 9, e56261. doi:10.7554/eLife.56261
- Jan-Matthis Lueckmann, Jan Boelts, David Greenberg, Pedro Goncalves, Jakob Macke: Benchmarking Simulation-Based Inference. Proceedings of The 24th International Conference on Artificial Intelligence and Statistics, PMLR 130:343-351, 2021.
- Yunfei Huang, Postdoc
- Tobias Machnitzki, PhD student
- Shivani Sharma, PhD student
- Andrey Vlasenko, Postdoc
- Vadim Zinchenko, Masters student
- Kubilay Demir, Masters student
- Marcel Nonnenmacher, Postdoc