Title :
Training ANN using linear minimax techniques
Author :
Charalambous, Chris
Author_Institution :
Dept. of Public & Bus. Adm., Cyprus Univ., Nicosia, Cyprus
Abstract :
The purpose of this paper is to present a new method for training ANN. The method solves a sequence of linear minimax optimization problems and does not make any assumption of the network structure, but it builds up as the algorithm proceeds. The method does not create unnecessary regions of local minima and it guarantees the classification of the input feature space in a finite number of steps
Keywords :
learning (artificial intelligence); minimax techniques; neural nets; pattern classification; classification; input feature space; learning; linear minimax techniques; neural networks; optimization; threshold weight; Linear programming; Minimax techniques; Neurons; Nonlinear equations; Optimization methods; Vectors;
Conference_Titel :
Neural Networks, 1995. Proceedings., IEEE International Conference on
Conference_Location :
Perth, WA
Print_ISBN :
0-7803-2768-3
DOI :
10.1109/ICNN.1995.487907
