Inversion of multilayer neural network with modelling error compensation

DOI10.1080/00207729708929442MaRDI QIDQ4355358zbMATH OpenOpenAlexFDO

Authors J.-H. Oh, Jeongwoo Lee

Publication date 17 September 1997

Published in International Journal of Systems Science. Principles and Applications of Systems and Integration (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1080/00207729708929442

zbMATH Keywords

feedforward structureinversion methoditerative Newton-Raphson schememultilayer neural network

Mathematics Subject Classification ID

Neural networks for/in biological studies, artificial life and related topics (92B20) General systems theory (93A99)

Review text

An inversion method for a multilayer neural network is developed in a systematic way, relying on the iterative Newton-Raphson scheme. The method mainly consists in the linearization of the neural network mapping, followed by the minimization of an appropriate cost function. To compensate the effect of the errors introduced by the inversion procedure, a feedforward structure is proposed; this solution is briefly compared with other principles that are currently used for compensation. Computer experiments demonstrate good numerical performances in solving inverse kinematics of redundant robots and controlling nonlinear plants.

Recommendations

• Model inversion using fuzzy neural network with boosting of the solution
• Neural network approach to computing matrix inversion
• Robot inverse kinematics via neural and neurofuzzy networks: architectural and computational aspects for improved performance
• Model inversion by parameter fit using NN emulating the forward model - evaluation of indirect measurements
• Neural networks of adaptive inverse control systems

Cites work

• Adaptive control of a class of nonlinear discrete-time systems using neural networks
• Adaptively controlling nonlinear continuous-time systems using multilayer neural networks
• Approximation by superpositions of a sigmoidal function
• scientific article; zbMATH DE number 50543 (Why is no real title available?)
• scientific article; zbMATH DE number 51537 (Why is no real title available?)
• Multilayer feedforward networks are universal approximators
• Neural networks for control systems - a survey

Cited in

(12)

• Model of neurocontrol of redundant systems
• Model inversion by parameter fit using NN emulating the forward model - evaluation of indirect measurements
• A fast algorithm for inversion of MLP networks in design problems
• Robot inverse kinematics via neural and neurofuzzy networks: architectural and computational aspects for improved performance
• Robust nonlinear inversion of wave-tilt data
• Model inversion using fuzzy neural network with boosting of the solution
• Multilayer Perceptron with Functional Inputs: an Inverse Regression Approach
• Neural network based models in the inversion of temperature vertical profiles from radiation data
• Advances in Neural Networks – ISNN 2005
• Polyhedral mixture of linear experts for many-to-one mapping inversion and multiple controllers
• Neural net backlash compensation with Hebbian tuning using dynamic inversion
• Semi-physical neural modeling for linear signal restoration

This page was built for publication: Inversion of multilayer neural network with modelling error compensation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4355358)