Title :
Performance analysis of complex-valued neural networks with stochastic resonance
Author :
Kaihatsu, Naoto ; Isokawa, Teijiro ; Matsui, Nobuyuki ; Nishimura, Haruhiko
Author_Institution :
Grad. Sch. of Eng., Univ. of Hyogo, Himeji, Japan
Abstract :
In this paper, we explore a complex-valued neural networks (NNs) with incorporating the mechanism of stochastic resonance (SR). SR is a phenomenon where weak periodic signals in the system can be detected by the presence of the noise. It is expected that the combination of complex-valued and SR mechanism will improve the performance of NNs, rather than NNs with either of them. The performances are evaluated through the problem of affine transformations in two-dimensional space. It is shown that complex-valued NN with SR could achieve more precise transformations rather than conventional real-valued NN and complex-valued NN.
Keywords :
affine transforms; complex networks; neural nets; performance evaluation; signal detection; stochastic processes; white noise; SR; afflne transformations; complex-valued neural networks; performance analysis; periodic signals; stochastic resonance; Artificial neural networks; Heuristic algorithms; Neurons; Noise; Stochastic resonance; Strontium; Training; Affine Transformation; Back-Propagation; Complex-valued Neural Networks; Stochastic Resonance;
Conference_Titel :
SICE Annual Conference 2010, Proceedings of
Conference_Location :
Taipei
Print_ISBN :
978-1-4244-7642-8
