Deep Learning of Dynamic Systems using System Identification Toolbox(TM)

TL;DR

This paper discusses recent enhancements to MATLAB's System Identification Toolbox, focusing on integrating deep learning architectures for nonlinear dynamic system modeling, including neural state-space models and auto-encoding features.

Contribution

It introduces new deep learning-based modeling capabilities and tools within the System Identification Toolbox, improving nonlinear and reduced-order system modeling.

Findings

Enhanced neural state-space models for nonlinear systems

Auto-encoding features for reduced-order modeling

Deeper integration with machine learning techniques

Abstract

MATLAB(R) releases over the last 3 years have witnessed a continuing growth in the dynamic modeling capabilities offered by the System Identification Toolbox(TM). The emphasis has been on integrating deep learning architectures and training techniques that facilitate the use of deep neural networks as building blocks of nonlinear models. The toolbox offers neural state-space models which can be extended with auto-encoding features that are particularly suited for reduced-order modeling of large systems. The toolbox contains several other enhancements that deepen its integration with the state-of-art machine learning techniques, leverage auto-differentiation features for state estimation, and enable a direct use of raw numeric matrices and timetables for training models.