Title :
Solution of Inverse Problems Using Multilayer Quaternion Neural Networks
Author_Institution :
Dept. of Electron. & Comput. Syst., Takushoku Univ., Hachioji, Japan
Abstract :
Neural network models extended to higher-dimensional numbers have been studied in recent years. In particular, quaternions have an advantage with respect to the expression of rotation in a three-dimensional space. On the other hand, the problem that estimates the cause of an observed result is called an inverse problem, whose solutions have been studied in various engineering fields. In this study, the quaternion network inversion is proposed as a neural network solution of the inverse problem extended to the quaternion. The simulation of the inverse mapping problem is examined to show the effectiveness of the proposed method.
Keywords :
inverse problems; learning (artificial intelligence); mathematics computing; neural nets; number theory; higher-dimensional numbers; inverse mapping problem; learning; multilayer quaternion neural networks; neural network model; neural network solution; quaternion network inversion; Aerospace electronics; Biological neural networks; Estimation; Inverse problems; Nonhomogeneous media; Quaternions; inverse problems; network inversion; neural networks; quaternion;
Conference_Titel :
Computational Science and Computational Intelligence (CSCI), 2014 International Conference on
Conference_Location :
Las Vegas, NV
DOI :
10.1109/CSCI.2014.152