一种基于PCA-BP神经网络的示例优选方法

在音频示例检索的研究中,针对示例数据量大而导致计算代价大、检索时间长和噪声鲁棒性差等问题,提出了一种基于主成分分析（PCA）和BP神经网络（BPNN）的示例优选方法。以信号鲁棒性评分为依据构建数据集合,使用主成分分析得到段级特征,消除数据冗余,减少输入变量,最后利用BPNN对保留成分进行建模预测。用PCA-BPNN模型对实验数据进行了验证性测试和分析,结果表明,该方法可以准确而高效地从一段音频中选取鲁棒性好的示例。     

遗传神经网络在基片图像边缘检测中的应用

由于复杂噪声信号的存在,使得对基片边缘的检测不理想.文章提出了一种利用遗传算法与BP神经网络相结合的图像边缘检测方法.先对基片图像进行中值滤波,将选中的样本归一化;后用遗传算法对神经网络权值进行优化:再用BP算法时神经网络进行修正,得到局部最优;最后利用此神经网络对舍有噪声信号的陶瓷基片进行边缘检测.实验表明,此种方法是有效的.     

基于PSO优化BPNN估计光伏阵列MPPT控制系统研究

为了充分利用光伏阵列转换能量,提高光伏阵列的发电效率,在分析光伏阵列的伏安特性及最大功率点跟踪(MPPT)原理的基础上,提出了一种基于粒子群算法优化BP神经网络(PSO-BPNN)的建模方法,并用这种改进的神经网络构建了光伏阵列的动态模型.通过PSO-BPNN模型拟合光伏阵列输出功率与输出电压的非线性关系,实现了对光伏阵列的最大功率点跟踪.Matlab/Simulink仿真及在线测试结果表明:基于PSO-BPNN估计的光伏阵列MPPT控制系统能快速、精确地跟踪光伏阵列的最大功率点,改善了BP神经网络收敛速度慢,易陷入局部极值,建模精度不高的缺点,提高了系统的稳定性和能量转换效率,是研究光伏发电这个复杂非线性系统的一个可行办法.     

一种提高云计算环境下检测入侵者速度和正确率的新方法

云计算环境要求入侵检测系统(IDS)极其快速和准确,用于云计算的智能型IDS——反向传播神经元网络(BPNN)经常出现泛化问题,即BPNN无解或总误差函数不能收敛于全局最小值.泛化问题降低了BPNN的识别速度和正确率.为了解决该问题,提出两种解决方法.第一种是剔除相关性大的那些特征,保留相关性较小或互相独立的重要特...     

SPP1 and the risk score model to improve the survival prediction of patients with hepatocellular carcinoma based on multiple algorithms and back propagation neural networks

Hepatocellular carcinoma (HCC) is associated with poor prognosis and fluctuations in immune status.Although studies have found that secreted phosphoprotein 1 (SPP1) is involved in HCC progression, its independentprognostic value and immune-mediated role remain unclear. Using The Cancer Genome Atlas and Gene ExpressionOmnibus data, we found that low expression of SPP1 is significantly associated with improved survival of HCCpatients and that SPP1 expression is correlated with clinical characteristics. Univariate and multivariate Coxregression confirmed that SPP1 is an independent prognostic factor of HCC. Subsequently, we found that T cell CD4memory-activated monocytes, M0 macrophages, and resting mast cells showed significant differences in penetrationin the high and low SPP1 expression groups. Next, we used the Weighted Gene Co-Expression Network and LeastAbsolute Shrinkage Sum Selection Operator algorithms to construct a risk score for the 9-immune-related genessignature. The risk score showed a good ability to identify high and low-risk patients and improved survivalprediction. We also used multivariate Cox regression to validate that risk score was significantly correlated with SPP1and overall survival. Lastly, the Back-Propagation Neural Network confirmed the reliability of the results of multiplealgorithms. In conclusion, the findings suggest that SPP1 is an independent marker of HCC survival and immunotherapy.     

Neural network approach to recognize the grid patterns in experimental mechanics

In this paper, we develop a neural network system that applies to the work of experimental mechanics. Two procedures are presented: one is the color calibration that used to filter 256 colors into three colors by adaptive-competitive learning network (ACLN). The other is the pattern recognition that is used to recognize the grid patterns and search the intersection points of grids by back propagation neural network (BPNN). In this paper, plane extrusion through a cosine-curved die is adopted to illustrate as example. The results show that the correlative points between the calculation and experiment are shown quite well agreement.     

改进GA-BP神经网络的无线信道预测方法

无线信道建模与仿真中,实现一种高效率、高准确性的无线信道预测方法是具有非常重要意义的。针对该需求,提出一种基于多种群遗传算法（Multi-population genetic algorithm, MPGA）-反向传播（Back propagation, BP）神经网络的无线信道预测方法。该方法通过改进遗传算法,优化BP神经网络中神经元的权值和阈值,以此改善BP神经网络预测精度差的问题,从而极大提高了BP神经网络的预测性能。将射线跟踪算法的理论值和BP神经网络结合,实现更高效的无线信道预测方法。通过对比遗传算法（Genetic algorithm, GA）-BP神经网络模型和MPGA-BP神经网络模型的预测误差,发现MPGA-BP神经网络模型的预测结果优于GA-BP神经网络模型,证明了所提出无线信道预测方法具有良好的精确度,可以更高效地进行无线信道预测。     

Quality of experience evaluation of HTTP video streaming based on user interactive behaviors

User interactive behaviors play a dual role during the hypertext transfer protocol (HTTP) video service: reflection and influence. However, they are seldom taken into account in practices. To this end, this paper puts forward the user interactive behaviors, as subjective factors of quality of experience (QoE) from viewer level, to structure a comprehensive multilayer evaluation model based on classic network quality of service (QoS) and application QoS. First, dual roles of user behaviors are studied and the characteristics are extracted where the user experience is correlated with user interactive behaviors. Furthermore, we categorize QoE factors into three dimensions and build the metric system. Then we perform the subjective tests and investigate the relationships among network path quality, user behaviors, and QoE. Ultimately, we employ the back propagation neural network (BPNN) to validate our analysis and model. Through the simulation experiment of mathematical and BPNN, the dual effects of user interaction behaviors on the reflection and influence of QoE in the video stream are analyzed, and the QoE metric system and evaluation model are established.     

基于数据挖掘的入侵检测多分类模型研究

入侵检测系统捕获的大量网络数据,需要采取有效的算法进行分类,用以判别数据流是否异常以及所属的攻击类别。为了有效地挖掘出判别数据流的模式与规则,针对入侵检测系统的分类器模型,采用数据挖掘技术中的反向传播神经网络算法、有监督的Kohonen神经网络算法和支持向量机算法进行对比研究。主要分析了各个算法的分类正确率、检测率和误...     

肌音信号模式识别在嵌入式实时手语翻译系统中的应用

基于肌音信号（Mechanomyogram, MMG）的模式识别是指采集MMG信号,应用机器学习算法进行动作识别的过程。为了实现手语实时分类,本文采用基于STM32芯片的轻量级嵌入式设备,控制双轴加速度传感器采集了前臂3块肌肉的6通道MMG,应用反向传播神经网络（Back propagation neural network, BPNN）算法建立分类模型,并将模型参数导入嵌入式系统中,实现算法的移植。实验结果表明该嵌入式系统可实现30种手语的实时识别,模型自测识别率达99.6%,实时识别中可达97.5%,每个动作分类所需时间少于0.52 ms,满足实时性要求,具有较高的实际应用价值。本文的研究结果可应用于人体康复工程,哑语翻译器,义肢控制等领域。