Paper:
Views over last 60 days: 374
Difference of Solution Regions due to Net Polarity
Hiromu Gotanda*, Hiroshi Shiratsuchi**, Katsuhiro Inoue**, and Kousuke Kumamaru
*Faculty of Engineering, Kinki University in Kyushu, 820 Iizuka, Japan
**Faculty of Computer Science and Systems Engineering, Kyushu Institute of Technology, 820 Iizuka, Japan
Received:May 23, 1997Accepted:August 1, 1997Published:October 20, 1997
Keywords:Neural net, XOR problem, Solution region, Attractor
Abstract
Solution for the XOR problem are formulated by normal vectors orthogonal to separation hyperplanes when nonlinear units are given by threshold functions. The shapes of solution regions are shown to be different for bipolar and unipolar nets. This suggests that attractors for BP leaning depend on polarity.
Cite this article as:H. Gotanda, H. Shiratsuchi, K. Inoue, and K. Kumamaru, “Difference of Solution Regions due to Net Polarity,” J. Robot. Mechatron., Vol.9 No.5, pp. 393-397, 1997.Data files:
