Levenberg-Marquardt算法入侵检测系统

摘要

In order to avoid lots of overlap space among mature detectors for random search strategy and slow convergence speed for evolutionary search method in real-valued detector generation algorithm, a co-evolution of the detection algorithm is proposed based on co-evolution algorithm mechanisms among the nature of species. The experimental tests demonstrate that the algorithm can not only achieve a more precise coverage of the nonself space with fewer detectors but also show a higher convergence speed.%本文针对传统BP神经网络算法学习速度慢、收敛性较差的问题,在Windows操作系统下,利用Levenberg-Marquardt算法进行改进,将优化后的LM算法运用到主机入侵检测中去,建立LMBP-HIDS入侵检测系统模型.实验结果表明,运用Levenberg-Marquardt优化算法进行主机入侵检测,改善了传统模型收敛速度慢、易陷入局部最小点、计算量大的缺点,可以较好地提高学习速率,缩短训练过程.