Title :
The functional link net in structural pattern recognition
Author :
Klaseen, Myungsook ; Pao, Y.-H.
Author_Institution :
Dept. of Inf. Eng., Chinese Univ. of Hong Kong, Shatin, Hong Kong
Abstract :
The functional link net is a higher-order generalized delta rule net. With it, input patterns are enhanced by means of functional transformations. In the simulations presented, the capabilities of the functional link net are explored by applying it to the structural pattern recognition, where the structural relationship between parts of a pattern that are of interest and the relevant parts of a pattern generally are not specified. Handwritten numerals are used as structured patterns. The tensor link is used to create higher order terms in the functional link net. The ideas of local templates, and a translation invariance are incorporated into the functional link net to reduce the higher-order terms
Keywords :
learning systems; neural nets; pattern recognition; functional link net; functional transformations; handwritten numerals; higher-order generalized delta rule net; learning systems; local templates; neural nets; structural pattern recognition; tensor link; translation invariance; Books; Collaboration; Computer vision; Field-flow fractionation; Logic; Neural networks; Pattern recognition; Tensile stress;
Conference_Titel :
Computer and Communication Systems, 1990. IEEE TENCON'90., 1990 IEEE Region 10 Conference on
Print_ISBN :
0-87942-556-3
DOI :
10.1109/TENCON.1990.152674
