搜索资源列表
hopfield_net
- 对于含有噪声的的数字图像模式“1、2、3、4”，使用“hopfield”神经网络，进行联想、识别，以去除噪声的影响。-for containing noise of the digital image model "1,2,3,4", "hopfield" neural network, Lenovo, identification, in order to remove noise impact.
bys
- 采用贝叶斯正则化算法提高BP网络的推广能力。在本例中，将采用两种训练方法，即L-M优化算法（trainlm)和贝叶斯正则化算法（trainbr),用以训练BP网络，使其能够拟合某一附加有白噪声的正弦样本数据。-The use of Bayesian regularization algorithm for BP network to improve generalization ability. In this case, two types of training methods will b
fpe
- This function calculates Akaike s final prediction error ％ estimate of the average generalization error. ％ ％ [FPE,deff,varest,H] = fpe(NetDef,W1,W2,PHI,Y,trparms) produces the ％ final prediction error estimate (fpe), the effective number of
nnfpe
- This function calculates Akaike s final prediction error ％ estimate of the average generalization error for network ％ models generated by NNARX, NNOE, NNARMAX1+2, or their recursive ％ counterparts. ％ ％ [FPE,deff,varest,H] = nnfpe(method,Ne
neuralnetwork
- 用于图像识别的新型神经元网络 提出了一种用于图像识别的映封正交神经元网络。 在一般情况下待识别的样本空间的维数相当大, 为了有效地进行识别,必须把样本空间的维数降下来。目前常用的方法就是特征提取法,本文采用映射正交过程把样本空间映射成正交分类空间,并在此基础上,采用网络进行图像分类。计算机上模拟结果表明此网络县有时缺损和噪声图像进行正确识别的能力-Image recognition for a new type of neural network used in a video im
CR
- Cognitive Radio systems with liscense and Rental Users in addtion to adding noise and attenuation.(Beautiful)-Cognitive Radio systems with liscense and Rental Users in addtion to adding noise and attenuation.(Beautiful)
zuselvfangzhen
- 移动通信中排队的阻塞率编程计算，运用C语言编写-Mobile Communications line of blocking probability calculation program, using C language. . . . .
solter_filter
- 摘要 本文提出了一种新的基于极大似然法的椒盐噪声滤波算法。在传统BP 算法中引入 了极大似然估计，在训练样本时能够在考虑网络逼近行为的同时对噪声分布进行估计。而且 针对椒盐噪声模型构造了新的鲁棒误差函数，从而使算法本身的抗干扰性增强。实验结果表 明了该算法与传统BP 算法相比具有更好的滤波性能。-Abstract This paper presents a new maximum likelihood method based on impulse noise filter. In
decisor_correlacao_generalizado
- It s a neural network decisor for Pareto set based in cross-correlation between a specific noise model and sucessive aproximation errors of the neural network that generates each solution of Pareto set.
ann
- 介绍了一种基于神经网络白化匹配滤波器的QRS 波检测方法。我们用神经网络白化匹配滤波器来处 理ECG 信号的低频成分, 模拟其非线性及非稳态的特性。处理后的信号中含有ECG 中大部分高频成分, 让其通过 一线性匹配滤波器来检测QRS 波及其位置。对于大噪声的ECG 信号, 在匹配滤波器后加差分滤波, 取平方及滑动 平均等处理, 提高检测正确率。使用这种方法我们对M IT?B IH 心电信号数据库中噪声比较大的105号数据进行的 处理, 检测正确率为9912 。作为对比, 用数字
modified_sinr
- The signal received by a user over the forward link in such a high data rate system contains interference from the neighboring base stations. we note that the strength of this interference is a function of the network load in the neighboring cells. w
modified_sinr_cal
- The signal received by a user over the forward link in such a high data rate system contains interference from the neighboring base stations. we note that the strength of this interference is a function of the network load in the neighboring cells. w
NoiseCancellation
- 仿真飞机上消噪耳机的工作流程，核心为基于神经网络的自适应滤波器-Simulation of headphone capable to cancel the noise in matlab with the help of adaptive filter based on neural network
443C_Simion_Alexandru
- The (a) section of Figure 1 illustrates the Line Echo Canceller (LEC) residing in the access network where echoes occur. Other voice processing modules, such as Automatic Level Control (ALC) and Noise Reduction (NR), are shown merely for illustration
Source-Localization-in-UWSAN
- 文章针对低信噪比下的水下目标定位问题，建立了水下无线传感器阵列网络，该结构包括多个分布式声传感器阵列，它适应于多模态信号处理，既可以利用目标的方位信息，又可以用能量信息。文中提出了用每个阵列接收到的信号能量作为参量完成目标定位并推导了基于能量的最大似然比目标定位方法。数值仿真表明：基于该结构的能量似然函数定位方法，可以有效估计目标的位置。并且比单阵元网络的定位性能和信息传输率上有了较大的提高， 尤其是在低信噪比下情况下，可以大大减小估计的方差。-With novel underwater wir
COPERATIVE-COMMUNICATIONS
- The channel is modeled by adding the Gaussian noise and Rayleigh Fading. Different combining methods and relaying protocols are studied and compared in this project. We analyzed performance characterizations in terms of The Bit Error Rate and the Thr
hopfield
- 一、 实验目的 1. 通过本实验加深对Hopfield神经网络原理的认识与理解； 2. 进一步熟练MATLAB软件和该软件平台上神经网络的模式识别功能。 二、 实验内容 本实验利用Hopfield神经网络来对0-9这10个数字进行识别，并实现噪声数字的识别。 -First, the purpose of the experiment Through this experiment to deepen knowledge and understanding of the H
vg-matlab
- 当前学界中已有的三种通过复杂网络理论分析时间序列的方法（VG、HVG、LPVG）都能够继承时间序列的基本特征，但相应性能例如抗噪以及算法复杂度等都有一 定的局限性 。因此，本文在已有三种方法的基础上提出另外两种可视图建网方法（LPHVG和DLPVG）。-Methods (VG, HVG, LPVG) already in the academic world through a complex network theory of three sequence analysis time are
mhktpmhv
- 是机器学习的例程，EOlqGVQ参数关于神经网络控制，包括调制，解调，信噪比计算，部分实现了追踪测速迭代松弛算法，tpSManb条件是一种双隐层反向传播神经网络，在matlab R2009b调试通过。- Machine learning routines, EOlqGVQ parameter On neural network control, Includes the modulation, demodulation, signal to noise ratio calculation, P
bpNeural-network-instance
- 例1 采用动量梯度下降算法训练 BP 网络。 例2 采用贝叶斯正则化算法提高 BP 网络的推广能力。在本例中，我们采用两种训练方法，即 L-M 优化算法（trainlm）和贝叶斯正则化算法（trainbr），用以训练 BP 网络，使其能够拟合某一附加有白噪声的正弦样本数据。-Example 1 uses the momentum gradient descent algorithm to train the BP network. Example 2 uses the Bayesian