Improved Competitive Hopfield Neural Network for Polygonal Approximation Problem
Zhou Ya-lan · 2009
Polygonal approximation is an important method to detect dominant points and represent curves.An Improved Competitive Hopfield Neural Network (ICHNN) was proposed for polygonal approximation problem.The proposed algorithm reduces research spaces by deleting the collinear points,redefines energy function of the network to represent the aim more efficiently,and combines with a merge-split method to help the network escape from the local mini-ma.The experimental results show that ICHNN is robuster and better than other algorithms such as dominant point detection,competitive Hopfield neural network, and chaotic Hopfield network,genetic algorithm.