On the effects of initialising a neural network with prior knowledge

Author

Andrews, Robert ; Geva, Shlomo

Author_Institution

Fac. of Inf. Technol., Queensland Univ., Brisbane, Qld., Australia

Volume

1

fYear

1999

fDate

1999

Firstpage

251

Abstract

This paper quantitatively examines the effects of initialising a Rapid Backprop Network (REP) with prior domain knowledge expressed in the form of propositional rules. The paper first describes the RBP network and then introduces the RULEIN algorithm which encodes propositional rules as the weights of the nodes of the REP network. A selection of datasets is used to compare networks that began learning from tabula rasa with those that were initialised with varying amounts of domain knowledge prior to the commencement of the learning phase. Network performance is compared in terms of time to converge, accuracy at convergence, and network size at convergence

Keywords

backpropagation; multilayer perceptrons; performance evaluation; radial basis function networks; RULEIN algorithm; Rapid Backprop Network; convergence; datasets; learning; neural network initialisation; neural network performance; prior knowledge; propositional rules; radial basis function network; three layer neural nets; Artificial neural networks; Australia; Convergence; Equations; Function approximation; Information technology; Learning systems; Machine learning; Neural networks; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on

Conference_Location

Perth, WA

Print_ISBN

0-7803-5871-6

Type

conf

DOI

10.1109/ICONIP.1999.843995

Filename

843995