分析了在奇异信号检测等问题中,使用M a llat算法对信号小波系数做单支重构波形存在失真的原因.通过调整M a llat算法的运算流程,得到了小波变换的一种基于共轭正交滤波器组的重构算法.理论分析与仿真试验结果表明,文中算法使上述失真...分析了在奇异信号检测等问题中,使用M a llat算法对信号小波系数做单支重构波形存在失真的原因.通过调整M a llat算法的运算流程,得到了小波变换的一种基于共轭正交滤波器组的重构算法.理论分析与仿真试验结果表明,文中算法使上述失真现象得到了改善,同时具有正交性、对称性和有限支撑性等优良性质,可用于对信号作精确的分析.展开更多
为了提高财务危机预测的性能,采用循环神经网络(Recurrent neural network,RNN)对关键指标进行分析训练,以解决因为时间变化带来的深度学习网络预测准确率性能下降的问题。选取关键指标特征生成预测样本,并充分利用RNN在时间序列的循环...为了提高财务危机预测的性能,采用循环神经网络(Recurrent neural network,RNN)对关键指标进行分析训练,以解决因为时间变化带来的深度学习网络预测准确率性能下降的问题。选取关键指标特征生成预测样本,并充分利用RNN在时间序列的循环计算优势,采用差异化时间序列的赋权策略,记忆不同历史时间序列对RNN预测分析的影响。经过RNN训练,并采用隐藏层输出不断循环的方式,将历史时间段输入不断作用于当前训练输出。引入人工蜂群(Artificial bee colony,ABC)算法在RNN反向传播过程中对时间序列权重进行更新。将RNN网络输出值与预测值的均方误差作为ABC的适用度函数,获得全局最优的ABC-RNN预测模型。试验证明,合理优化历史时间序列权重,能够获得较高的危机预测准确率。和常用预测算法对比,所提ABC-RNN算法的预测准确率更高,且曲线面积(Area under the curve,AUC)值更高。展开更多
According to the operational conditions of an aviation module reticule,a measurement mode is proposed,which is based on an industrial photogrammetry system,with matching by a measuring pen.Meanwhile,the factors affect...According to the operational conditions of an aviation module reticule,a measurement mode is proposed,which is based on an industrial photogrammetry system,with matching by a measuring pen.Meanwhile,the factors affecting the accuracy of the measurement have been analyzed and verified by examples.The analysis is described as follows:①Along the optical axis of the camera,the error is larger than the ones in other directions using the“single camera+measuring pen”mode;②By avoiding the error along the optical axis of the camera,the accuracy of the“single camera+measuring pen”mode is better than 0.1 mm when the measuring pen is moving parallel to the optical axis.展开更多
文摘分析了在奇异信号检测等问题中,使用M a llat算法对信号小波系数做单支重构波形存在失真的原因.通过调整M a llat算法的运算流程,得到了小波变换的一种基于共轭正交滤波器组的重构算法.理论分析与仿真试验结果表明,文中算法使上述失真现象得到了改善,同时具有正交性、对称性和有限支撑性等优良性质,可用于对信号作精确的分析.
文摘为了提高财务危机预测的性能,采用循环神经网络(Recurrent neural network,RNN)对关键指标进行分析训练,以解决因为时间变化带来的深度学习网络预测准确率性能下降的问题。选取关键指标特征生成预测样本,并充分利用RNN在时间序列的循环计算优势,采用差异化时间序列的赋权策略,记忆不同历史时间序列对RNN预测分析的影响。经过RNN训练,并采用隐藏层输出不断循环的方式,将历史时间段输入不断作用于当前训练输出。引入人工蜂群(Artificial bee colony,ABC)算法在RNN反向传播过程中对时间序列权重进行更新。将RNN网络输出值与预测值的均方误差作为ABC的适用度函数,获得全局最优的ABC-RNN预测模型。试验证明,合理优化历史时间序列权重,能够获得较高的危机预测准确率。和常用预测算法对比,所提ABC-RNN算法的预测准确率更高,且曲线面积(Area under the curve,AUC)值更高。
文摘According to the operational conditions of an aviation module reticule,a measurement mode is proposed,which is based on an industrial photogrammetry system,with matching by a measuring pen.Meanwhile,the factors affecting the accuracy of the measurement have been analyzed and verified by examples.The analysis is described as follows:①Along the optical axis of the camera,the error is larger than the ones in other directions using the“single camera+measuring pen”mode;②By avoiding the error along the optical axis of the camera,the accuracy of the“single camera+measuring pen”mode is better than 0.1 mm when the measuring pen is moving parallel to the optical axis.
