基于人工神经网络降维映射的统计优化方法

A Statistic Optimization Method Based on Neural Network Reducing Dimensional

提出将多维空间的样本数据降维映射到二维平面上，并在该平面上自动生成函数的等值分布曲线，从而，可直观出该函数的最优点或最优区域，通过本文提出的逆映射算法可将其还原到多维空间用原始变量表示．运算实例结果表明，基于神经网络降维映射的优化方法，直观、准确、可靠，对于有约束优化问题的求解特别有效．

Sample datas in multidimensinal space is mapped and reduced dimension to a two dimensional plane by a neural network, and the contours of fuction are produced automaticlly on the plane. The optimal operation point or region can be determined intuitively based on this mapping plane, by means of inversion mapping algorthm presented this paper, the optimal point or region is mapped inversely to original multidimensional and is can represented with operation variables. Example research shows that this statistic optimization method is intuition, accuracy and reliability, especialy, has very good effect in solving constrained optimization problems.