A hybrid objective function for robustness of artificial neural networks -- estimation of parameters in a mechanical system

DOI10.1553/ETNA_VOL56S209MaRDI QIDQ2672200zbMATH OpenOpenAlexFDO

Authors Udo Schröder, Jan Sokołowski, Volker Schulz, H.-P. Beise

Publication date 8 June 2022

Published in ETNA - Electronic Transactions on Numerical Analysis (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2004.07692

IPFS File on IPFS (Hint: this is only the Hash - if you get a timeout, this file is not available on our server.)

zbMATH Keywords

convolutional neural networksdynamical systemsmathematical modellingparameter estimationprediction robustnesssequential datasystem identification

Mathematics Subject Classification ID

Networked control (93B70) Estimation and detection in stochastic control theory (93E10) Identification in stochastic control theory (93E12)

Abstract: In several studies, hybrid neural networks have proven to be more robust against noisy input data compared to plain data driven neural networks. We consider the task of estimating parameters of a mechanical vehicle model based on acceleration profiles. We introduce a convolutional neural network architecture that is capable to predict the parameters for a family of vehicle models that differ in the unknown parameters. We introduce a convolutional neural network architecture that given sequential data predicts the parameters of the underlying data's dynamics. This network is trained with two objective functions. The first one constitutes a more naive approach that assumes that the true parameters are known. The second objective incorporates the knowledge of the underlying dynamics and is therefore considered as hybrid approach. We show that in terms of robustness, the latter outperforms the first objective on noisy input data.

Recommendations

• Robust estimation of parametric membership regions
• On robust training of regression neural networks
• Probabilistic robustness estimates for feed-forward neural networks
• Achieving a more robust neural network model for control of a MR damper by signal sensitivity analysis
• scientific article; zbMATH DE number 2111322

Cites work

• An Introduction to Variational Autoencoders
• Data driven governing equations approximation using deep neural networks
• Data-driven identification of parametric partial differential equations
• Deep learning of dynamics and signal-noise decomposition with time-stepping constraints
• Geometric Numerical Integration
• Neural network architectures for parameter estimation of dynamical systems
• Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
• System identification. An introduction.

Cited in

(1)

• Establishing criteria to ensure successful feedforward artificial neural network modelling of mechanical systems

Describes a project that uses

Uses Software

• AlexNet
• ImageNet
• BERT
• torchdiffeq
• word2vec
• MNIST

This page was built for publication: A hybrid objective function for robustness of artificial neural networks -- estimation of parameters in a mechanical system

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2672200)