Mini-batch optimization enables training of ODE models on large-scale datasets.

Autor: Stapor P; Helmholtz Zentrum München - German Research Center for Environmental Health, Institute of Computational Biology, 85764, Neuherberg, Germany.; Technische Universität München, Center for Mathematics, Chair of Mathematical Modeling of Biological Systems, 85748, Garching, Germany., Schmiester L; Helmholtz Zentrum München - German Research Center for Environmental Health, Institute of Computational Biology, 85764, Neuherberg, Germany.; Technische Universität München, Center for Mathematics, Chair of Mathematical Modeling of Biological Systems, 85748, Garching, Germany., Wierling C; Alacris Theranostics GmbH, 12489, Berlin, Germany., Merkt S; Universität Bonn, Faculty of Mathematics and Natural Sciences, 53115, Bonn, Germany., Pathirana D; Universität Bonn, Faculty of Mathematics and Natural Sciences, 53115, Bonn, Germany., Lange BMH; Alacris Theranostics GmbH, 12489, Berlin, Germany., Weindl D; Helmholtz Zentrum München - German Research Center for Environmental Health, Institute of Computational Biology, 85764, Neuherberg, Germany., Hasenauer J; Helmholtz Zentrum München - German Research Center for Environmental Health, Institute of Computational Biology, 85764, Neuherberg, Germany. jan.hasenauer@uni-bonn.de.; Technische Universität München, Center for Mathematics, Chair of Mathematical Modeling of Biological Systems, 85748, Garching, Germany. jan.hasenauer@uni-bonn.de.; Universität Bonn, Faculty of Mathematics and Natural Sciences, 53115, Bonn, Germany. jan.hasenauer@uni-bonn.de.
Jazyk: angličtina
Zdroj: Nature communications [Nat Commun] 2022 Jan 10; Vol. 13 (1), pp. 34. Date of Electronic Publication: 2022 Jan 10.
DOI: 10.1038/s41467-021-27374-6
Abstrakt: Quantitative dynamic models are widely used to study cellular signal processing. A critical step in modelling is the estimation of unknown model parameters from experimental data. As model sizes and datasets are steadily growing, established parameter optimization approaches for mechanistic models become computationally extremely challenging. Mini-batch optimization methods, as employed in deep learning, have better scaling properties. In this work, we adapt, apply, and benchmark mini-batch optimization for ordinary differential equation (ODE) models, thereby establishing a direct link between dynamic modelling and machine learning. On our main application example, a large-scale model of cancer signaling, we benchmark mini-batch optimization against established methods, achieving better optimization results and reducing computation by more than an order of magnitude. We expect that our work will serve as a first step towards mini-batch optimization tailored to ODE models and enable modelling of even larger and more complex systems than what is currently possible.
(© 2022. The Author(s).)
Databáze: MEDLINE
Externí odkaz: