Partial observations and conservation laws: Grey-box modeling in biotechnology and optogenetics

arXiv1909.04234MaRDI QIDQ6324985FDOQ6324985

Authors: Robert J. Lovelett, Jose L Avalos, I. G. Kevrekidis 

Publication date: 9 September 2019

Abstract: Developing accurate dynamical system models from physical insight or data can be impeded when only partial observations of the system state are available. Here, we combine conservation laws used in physics and engineering with artificial neural networks to construct "grey-box" system models that make accurate predictions even with limited information. These models use a time delay embedding (c.f., Takens embedding theorem) to reconstruct effect of the intrinsic states, and can be used for multiscale systems where macroscopic balance equations depend on unmeasured micro/meso scale phenomena. By incorporating physics knowledge into the neural network architecture, we regularize variables and may train the model more accurately on smaller data sets than black-box neural network models. We present numerical examples from biotechnology, including a continuous bioreactor actuated using light through optogenetics (an emerging technology in synthetic biology) where the effect of unmeasured intracellular information is recovered from the histories of the measured macroscopic variables.

This page was built for publication: Partial observations and conservation laws: Grey-box modeling in biotechnology and optogenetics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6324985)