逆变电源的原理和应用.pdf

OpenGL版 胡事民 刘利刚 刘永进 张松海编译 清华大学出版社

thinkcmf微信登录插件，放在插件文件夹下，就可以直接安装。然后配置之后就可以直接调用

matlab编写的GPS和惯导的组合导航，内含仿真程序和真实的数据，大家可以下载下来仿真并得到结果。很好用的

包内包含主元分析算法（PCA）的Matlab完整故障诊断程序，附详细的word文档和PPT介绍

有深度学习中必读经典，以及相应的matlab代码。 此外，文章中本人做的笔记，希望能帮助大家更好的理解。文章为：1.A fast learning algorithm for deep belief nets（Hinton） 2. Learning Deep Architectures for AI （Bengio） 3. A Practical Guide to Training Restricted Boltzmann Machines（Hinton） 等。。code 为经典的deep learning tool（matlab版），有DB

Opencv下利用SIFT、SURF、ORB三种特征点实现图像匹配

针对粒子群优化进入后期得到解的精度低，不稳定的问题，提出了一种改进的自适应的算法，采用惯性权重线性减小，对成功的加速系数产生记忆，并于下一代更新中以一定概率唤醒记忆，通过仿真实验，表明可以有效的解决了后期精度差，不稳定的问题。

【实例简介】2013北航应用数理统计第一次大作业-回归分析

用于非线性最小二乘问题，通过高斯牛顿迭代实现

pdf文档，中文来源：华北电力大学学报242007[11997199811.2199910.88408200.68000.47800.2掩76072007000100120时间/天68040100120图3训练过程中的误差变化时间/天Fig.3 Errors during training图1日负荷预测时洲练数据4Fig. 1 Training data for day load for ecas ting35←只51.5050-10203040如0的070809010101200.8时间压图4训练过程中神经元个数变化Fig 4 Grow th of neurons during train ing0.200.95系0.940.8505101520253035滞后时间/天如0.80.75图2训练数据的自相关系数Fig. 2 Autocorrelation coefficients of training dat a(a拟合曲线0.〔(3)10.040.0系70/14060[1000000]时间/天(b拟合误差「O1000001图5训练结果及误差Fig. 5 Result and error of train in3- EBF6,C1994-2012cHinaAcademicJournalElectronicPublishingHouse.Allrightsreservedhttp://www.cnki.net4EUNITE2582080078076074072026.26700(3)66005101520253035时间/天50/1[00001图6日最大负荷预测结果[01010]26ig. 6 Result of forFrecasting g1 SOFNN[716.2,739.7,757.7,781.3],7Tab 1 Accuracy of training and forecast ing[720.1,738.2,763.7,MethodMAPE/ (%O) ME767.7],7SOFNN1.3344.137901.7850.04780EUNITE1.95E770760winner750报告中ME值没有准确堤供,但可以从报告中估轵740计得出约50~60完整的 EUNITE网络竟赛原始数据可730从EUNITE网站获得(http://neuron.tuke.sk/compe-720tit ion/ index. php)7103354EUNITEMAPE时间1.95[2]图7周平均最大负荷预测结果3.2Fig. 7 Forecasting result of week average max loadSOFNNSOFNNWSj72SOF NNWLWLideltaWli, delta WLi=WLj-Wli(3h:× delt al+k× deltal wj,Yh=0.58o1994-2012ChinaAcademicJournalElectronicPublishingHouse.Allrightsreservedhttp:/www.cnki.net262007820[1] Chen B J, Chang M W, Lin C J. Load forecasting using800support vector machines: a study on eunitE com-780tion 2001[J]. IEEE transactions on power systems7602004.19(4):1821-1830[2] Company behind East- Slov akia Power Distrib ution740Com panWorle w ide com petition w ithin the EU720NITE network, EUNITE competiton report [RI700[3]6802004.28(17):1-1105101520253035[4 Leng g, Prasad g, McGinnity T M. An on line algor-时间/天rithm for creating self organ izing fuzzy neural networ ks图8修正后的日最大负荷预测结果Neural Netw orks, 2004,(17): 1477-1493[5 Ort iz: Arroyo D, Skov M K, Huynh Q. Accurate Elee-F ig.8 Forecasting results after rev sedtricity Load Forecasting with Artificial Neur al NetworksIC. Proceedings of the 2005 International Conference2SOFNNon Compu tat io nal Intel ligence for M odel ling, Control andTah 2 Accuracy of forecast ingAuto mation. and International Conference on I ntelligentMethodMAPE/(%)MEAgents, Web Technolo gies and Internet CommerceSOFNN1.7850.04(CIMCAIA WTIC 05). 20051.5941.95[ 6] Tao X. Input dimens ion reduction for load forecastingEUNITE1.9550-60based on support vector machines [C]. Hong Kong82004 IEEE International Conference of Electric U til yderegulation, res tructur ing and pow er technolog iesMAPE ME20[7 Hsu CC. Dynam icall y Optim izing Parameters in SupportVector Regression An A pp licat io n of Electricity Load4结论Forecasting [C]. Haw aii: Proceedings of the 3 9thIlaw aii International Conference on System Sciences199918 Pan kratz A. Forecasting w ith Univariate Box-JenkinsModels [m. John Wiley sons, 1983SOFNN(1972-),,C1994-2012cHinaAcademicJournalElectronicPublishingHouse.Allrightsreservedhttp://www.cnki.net