战场目标价值评判的神经网络方法

在建立目标价值评判指标体系的基础上,利用BP神经网络构建3层结构多节点模型进行战场目标的价值排序.网络以对数sigmod函数为传递函数,选取典型目标组成学习和测试样本,采用改进后的弹性BP算法完成模型的训练和测试,由此绘制出网络误差变化曲线和权值、阈值Hinton图.经学习、测试和大量泛化目标的检测,模型能成功运用于战场目标价值排序.

Nerve Network Method for Battle Target Value Ordering

Based on evaluation index system of target value, set up the model to evaluating battle target value by BP nerve network, which had three layers and multi-cells. The network adopts logarithm sigmod function as transfer function. Select typical targets as learning and training stylebook. Adopt improved flexible BP algorithm to realize model training and testing, and draw the network error varied curve and weighting, threshold Hinton figure. After learning, testing and detecting for a great deal of general target, the model can be used in evaluating battle target value successfully.