How Learning Methods Influence the Performance of Complex-Valued Multilayer Perceptrons

DOI

Bibliographic Information

Other Title
  • 複素多層パーセプトロンの性能と学習法の関係に関する実験評価

Search this article

Description

Complex-valued multilayer perceptrons (C-MLPs) are expected to work well on the processing of signals such as radio waves and sound waves, which can be naturally expressed as complex numbers, since C-MLPs can naturally treat complex numbers. The performance of C-MLPs may seriously depend on learning methods because in the search space there exist many local minima and singular regions, which prevent learning methods from finding excellent solutions. C-BP and C-BFGS are well-known methods for learning C-MLPs. Moreover, complex-valued singularity stairs following (C-SSF) has recently been proposed, which achieves successive learning by utilizing singular regions and guarantees monotonic decrease of training errors. This paper evaluates how learning methods influence the performance of C-MLPs by doing experiments using three learning methods and eight benchmark datasets.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top