Title :
A study on the partitioning capabilities of two-layer neural networks
Author :
Lin, Che-Chern ; El-Jaroudi, Amro
Author_Institution :
Dept. of Electr. Eng., Pittsburgh Univ., PA, USA
fDate :
27 Jun-2 Jul 1994
Abstract :
Recent studies indicate that neural networks with a single hidden layer can implement both nonconvex and disconnected decision regions. In this paper, we examine the capabilities of neural nets to implement more complex decision regions such as nested decision regions. We study two-class problems with four, five and six nested decision regions and demonstrate that single hidden-layer nets can in fact implement these complex structures
Keywords :
neural nets; complex decision regions; disconnected decision regions; nested decision regions; partitioning capabilities; single hidden layer; two-class problems; two-layer neural networks; Convergence; Gradient methods; Measurement standards; Neural networks; Region 3; Region 4; Shape;
Conference_Titel :
Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-1901-X
DOI :
10.1109/ICNN.1994.374190
