针对BP网络现有学习算法的不足,在研究多种改进算法的基础上,在非线性规划理论的指导下,在共轭梯度法中巧妙地运行间插步骤和不精确的一维搜索技术,因而形成一种自适应的、具有全局收敛性的、快速的学习算法 Aiming at the deficiency of the existing learning algorithms of BP network, based on the research of many kinds of improved algorithms and under the guidance of nonlinear programming theory, the interlacing steps and inaccurate one-dimensional search technique are skillfully operated in the conjugate gradient method, Thus forming an adaptive, globally convergent and fast learning algorithm