Title :
Quaternion network inversion for solving inverse problems
Author :
Iura, Tsubasa ; Ogawa, Takehiko
Author_Institution :
Grad. Sch. of Eng., Takushoku Univ., Tokyo, Japan
Abstract :
Recently, the neural network model for handling the high dimensional numbers including a complex number has been proposed and studied in the field of artificial neural networks. Especially, the quaternion is focused in a number of fields such as computer graphics, because we easily express the algebra in three-dimensional space using quaternion. On the other hand, the problem which searches for a cause from an observed result is called an inverse problem, and has been studied in various fields. The network inversion has been studied as a neural network based solution for inverse problems. Recently, the extension to complex-valued neural network has been proposed. In this study, we propose the extension of the network inversion method to the quaternion neural networks.
Keywords :
inverse problems; neural nets; process algebra; algebra; artificial neural networks; complex number; computer graphics; high dimensional numbers; inverse problems; neural network model; quaternion network inversion; quaternion neural networks; three-dimensional space; Estimation; Inverse problems; Neural networks; Neurons; Nonhomogeneous media; Quaternions; Training; Network inversion; inverse problems; quaternion;
Conference_Titel :
SICE Annual Conference (SICE), 2012 Proceedings of
Conference_Location :
Akita
Print_ISBN :
978-1-4673-2259-1