[数值算法/人工智能] 供网上共享
面向语音帕金森病识别的分包融合算法
基于语音数据挖掘实现帕金森病诊断的方法近年来已被证明有效。然而，受数据采集对象患病程度以及采集设备和环境等因素影响，所获取数据集的样本空间中存在不同类别样本混叠现象。混叠区域的样本难以有效识别，严重影响了算法的分类准确度。为了解决这一问题，本文提出了分包融合集成算法，通过设计类心距离比值来衡量样本的混叠程度并将训练集划分成多个子集，再利用错误分类样本传递式训练的方法调整子集划分结果，最后通过优化子分类权重对各个子分类器的测试结果进行加权融合。实验结果表明，本文方法分类准确度在两个公共数据集上都得到明显提高，平均准确度最大提高可达25.44%。该方法不仅有效提高了帕金森病语音数据集分类准确度，还增加了样本利用率，为帕金森病语音诊断提供了一种新思路。
Packet fusion algorithm for speech recognition of Parkinson's disease (2020-09-06, matlab, 1386KB, 下载1次)
[数值算法/人工智能] elm分类代码
预测分类，很赞
ELM在研究中被视为一类特殊的FNN，或对FNN及其反向传播算法的改进，其特点是隐含层节点的权重为随机或人为给定的，且不需要更新，学习过程仅计算输出权重 [2]? 。 传统的ELM具有单隐含层，在与其它浅层学习系统，例如单层感知机（single layer perceptron）和支持向量机（Support Vector Machine, SVM）相比较时，被认为在学习速率和泛化能力方面可能具有优势 ? 。ELM的一些改进版本通过引入自编码器构筑或堆叠隐含层获得了深度结构，能够进行表征学习 [3]? 。 ELM的应用包括计算机视觉和生物信息学，也被应用于一些地球科学、环境科学中的回归问题
Elm is regarded as a special kind of FNN in the research, or the improvement of FNN and its back propagation algorithm. Its characteristic is that the weights of hidden layer nodes are random or given artificially, and do not need to be updated. The learning process only calculates the output weights [2]. Traditional elm has a single hidden layer. Compared with other shallow learning systems, such as single layer perceptron and support vector machine, it is considered to have advantages in learning speed and generalization ability. Some improved versions of elm obtain deep structure by introducing self encoder to build or stack hidden layer, which can be used for representation learning. The application of elm includes computer vision and bioinformatics, and it is also applied to some regression problems in Earth Science and environmental science (2020-05-29, matlab, 2KB, 下载4次)
[数值算法/人工智能] Kuschner-BayesianNetwork-feature
Kuschner论文，贝叶斯网络方法在质谱数据特征选择。其中关于机器学习中贝叶斯分类器部分有完整原理分析，可以用于认知无线电网络的频谱感知等新领域。含有matlab程序大于100页，子函数很多。
Kuschner paper, Bayesian network methods of feature selection in mass spectrometry data. One of the Bayes classifier machine learning part of a complete theory analysis, can be used for spectrum sensing cognitive radio networks and other new fields. Matlab program contains more than 100 pages, Functions a lot. (2011-10-29, matlab, 2013KB, 下载79次)
[数值算法/人工智能] matlab
MATLAB从入门到精通（配套光盘）
本书以MATLAB 7.x软件为基础，系统讲解了MATLAB基本环境和操作方法；分章阐述了矩阵计算、数值计算、符号计算、数据可视化、数据分析、M文件编写、 Simulink仿真、句柄图形、图形用户界面、文件读写、MATLAB编译器、应用程序接口等内容；并结合相关的函数或命令，精心编写了一些具体的示例，充分说明了具体函数和命令的使用方法。
本书所带的光盘提供全部示例的源代码，可帮助读者更好地理解书中的内容并更快地掌握MATLAB的使用方法。
本书内容充实、篇幅紧凑，既可作为高校学生系统学习MATLAB的书籍，也可以作为广大科研和工程技术人员在工作中使用MATLAB的参考书。
matlab (2009-11-01, matlab, 1150KB, 下载65次)
[数值算法/人工智能] pid_zen_ding
我们知道，PID控制器各参数对系统的影响是；增大开环比例系数 ，一般将加快系统的影响速度，在有静差的情况下则有利于减小静差；但过大的比例系数又会加大系统超调，甚至产生振荡，使系统不稳定。
We know that, PID controller parameters on the system is increasing the proportion of open-loop coefficients, the general will speed up the system speed, in the case of static error it will help reduce the static error but the proportion is too large coefficient will increase the system overshoot, even oscillation, allowing the system to instability. (2009-01-06, matlab, 10KB, 下载22次)
[数值算法/人工智能] gaf_pid2
我们知道，PID控制器各参数对系统的影响是；增大开环比例系数 ，一般将加快系统的影响速度，在有静差的情况下则有利于减小静差；但过大的比例系数又会加大系统超调，甚至产生振荡，使系统不稳定。
We know that, PID controller parameters on the system is increasing the proportion of open-loop coefficients, the general will speed up the system speed, in the case of static error it will help reduce the static error but the proportion is too large coefficient will increase the system overshoot, even oscillation, allowing the system to instability. (2009-01-06, matlab, 13KB, 下载4次)
[数值算法/人工智能] gaf_pid
我们知道，PID控制器各参数对系统的影响是；增大开环比例系数 ，一般将加快系统的影响速度，在有静差的情况下则有利于减小静差；但过大的比例系数又会加大系统超调，甚至产生振荡，使系统不稳定。
We know that, PID controller parameters on the system is increasing the proportion of open-loop coefficients, the general will speed up the system speed, in the case of static error it will help reduce the static error but the proportion is too large coefficient will increase the system overshoot, even oscillation, allowing the system to instability. (2009-01-06, matlab, 1KB, 下载3次)
[数值算法/人工智能] ekf_pf
基于卡尔曼滤波和粒子滤波器级联模型的静基座惯导初始对准算法及仿真
Based on the Kalman filter and particle filter cascade model static base INS initial alignment algorithm and simulation (2008-10-03, matlab, 212KB, 下载197次)
[数值算法/人工智能] gawnn
遗传算法优化小波神经网络的源程序:
1.构造的非线性函数: 位于nninit_test.m
2.直接用WNN逼近非线性:Wnn_test.m, (内部调用小波函数)
3.遗传算法优化后逼近 :GA_Wnn_test.m (内部调用遗传算法的,初始化,适应度,解码函数)
genetic algorithm optimization WNN source : 1. Construction of the nonlinear function : nninit_test.m at 2. WNN directly with nonlinear approximation : Wnn_test.m. (internal called wavelet function) 3. Genetic Algorithm optimization approach : GA_Wnn_test.m (internal called genetic algorithms, initialize, fitness and decoding functions) (2006-11-19, matlab, 7KB, 下载711次)