论文部分内容阅读
提出一种新的基于约束学习神经网络的递推分块方法,来分批(块)求解任意高阶多项式的任意数(小于多项式的阶)个根(包括复根).同时给出了基于多项式中根与系数间的约束关系构造的用于求根的BP网络约束学习算法,提出了对应的学习参数的自适应选择方法.实验结果表明,这种分块神经求根方法,相对传统方法,能够快速有效地获得任意高阶多项式对应的根.
A new recursive block method based on constrained learning neural networks is proposed to solve any arbitrary number (less than the degree of polynomial) of any higher order polynomials (including complex roots) in batches (blocks) The constrained learning algorithm based on BP network constructed by the constraint relation between roots and coefficients in polynomials is proposed, and the adaptive selection method of corresponding learning parameters is proposed. The experimental results show that this method of subdivision neural root, compared with the traditional method, It can quickly and efficiently obtain the root corresponding to any higher-order polynomial.