Relationship between phase and amplitude generalization errors in complex- and real-valued feedforward neural networks

Akira Hirose, Shotaro Yoshida

Research output: Contribution to journalArticlepeer-review

9 Citations (Scopus)

Abstract

We compare the generalization characteristics of complex-valued and real-valued feedforward neural networks. We assume a task of function approximation with phase shift and/or amplitude change in signals having various coherence. Experiments demonstrate that complex-valued neural networks show smaller generalization error than real-valued networks having doubled input and output neurons in particular when the signals have high coherence, that is, high degree of wave nature. We also investigate the relationship between amplitude and phase errors. It is found in real-valued networks that abrupt change in amplitude is often accompanied by steep change in phase, which is a consequence of local minima in real-valued supervised learning.

Original languageEnglish
Pages (from-to)1357-1366
Number of pages10
JournalNeural Computing and Applications
Volume22
Issue number7-8
DOIs
Publication statusPublished - 2013 Jun 1
Externally publishedYes

Keywords

  • Complex-valued neural network
  • Function approximation
  • Generalization

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence

Fingerprint Dive into the research topics of 'Relationship between phase and amplitude generalization errors in complex- and real-valued feedforward neural networks'. Together they form a unique fingerprint.

Cite this