Title :
Using stereographic projection as a preprocessing technique for Upstart
Author :
Saffery, James ; Thornton, Chris
Author_Institution :
Dept. of Artificial Intelligence, Edinburgh Univ., UK
Abstract :
Upstart is a constructive algorithm which learns by constructing a network of units rather than by modifying the weights of connections in a pre-wired network. The algorithm is not guaranteed to converge unless the input patterns satisfy certain criteria. However, this restriction can be overcome in some problems by using stereographic projection as a preprocessing technique. The technique is described, and an example is given in which the technique is used to enable Upstart to solve the two-spirals problem
Keywords :
convergence; neural nets; Upstart; constructive algorithm; convergence; input patterns; learning; network construction; preprocessing technique; stereographic projection; two-spirals problem; Artificial intelligence; Artificial neural networks; Binary trees; Bridges; Convergence; Error correction; Pattern classification;
Conference_Titel :
Neural Networks, 1991., IJCNN-91-Seattle International Joint Conference on
Conference_Location :
Seattle, WA
Print_ISBN :
0-7803-0164-1
DOI :
10.1109/IJCNN.1991.155373
