Statistical analysis of multilayer perceptrons performances

Author

Brad, Remus ; Mihu, Ioan ; Breazu, Macarie

Author_Institution

Comput. Sci. Dept, Lucian Blaga Univ. of Sibiu, Romania

Volume

4

fYear

2001

fDate

2001

Firstpage

2794

Abstract

The paper is based on a series of studies on the learning capabilities of multilayered perceptrons (MLP). The complexity of these nonlinear systems can be varied, acting for instance on the number of hidden units, but we will be confronted with a choice dilemma, concerning the optimal complexity of the system for a given problem. By the mean of statistical methods, we have found that the effective number of hidden units is smaller than the potential size; some units have a “binary” activation level or a time constant activation. We also prove that weight initialization to small values is recommended and reduce the effective size of the hidden layer

Keywords

computational complexity; learning (artificial intelligence); multilayer perceptrons; optimisation; statistical analysis; MLP; binary activation level; multilayer perceptron performances; optimal complexity; statistical analysis; time constant activation; weight initialization; Complexity theory; Computer science; Gaussian noise; Information retrieval; Learning systems; Multilayer perceptrons; Neural networks; Nonlinear systems; Statistical analysis; Transfer functions;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 2001. Proceedings. IJCNN '01. International Joint Conference on

Conference_Location

Washington, DC

ISSN

1098-7576

Print_ISBN

0-7803-7044-9

Type

conf

DOI

10.1109/IJCNN.2001.938816

Filename

938816