1、Good is good, but better carries it.精益求精，善益求善。MATLAB神經(jīng)網(wǎng)絡(luò)工具箱函數(shù)-MATLAB神經(jīng)網(wǎng)絡(luò)工具箱函數(shù)說(shuō)明：本文檔中所列出的函數(shù)適用于MATLAB5.3以上版本，為了簡(jiǎn)明起見(jiàn)，只列出了函數(shù)名，若需要進(jìn)一步的說(shuō)明，請(qǐng)參閱MATLAB的幫助文檔。1.網(wǎng)絡(luò)創(chuàng)建函數(shù)newp創(chuàng)建感知器網(wǎng)絡(luò)newlind設(shè)計(jì)一線(xiàn)性層newlin創(chuàng)建一線(xiàn)性層newff創(chuàng)建一前饋BP網(wǎng)絡(luò)newcf創(chuàng)建一多層前饋BP網(wǎng)絡(luò)newfftd創(chuàng)建一前饋輸入延遲BP網(wǎng)絡(luò)newrb設(shè)計(jì)一徑向基網(wǎng)絡(luò)newrbe設(shè)計(jì)一嚴(yán)格的徑向基網(wǎng)絡(luò)newgrnn設(shè)計(jì)一廣義回歸神經(jīng)網(wǎng)絡(luò)newpnn設(shè)計(jì)一概率

2、神經(jīng)網(wǎng)絡(luò)newc創(chuàng)建一競(jìng)爭(zhēng)層newsom創(chuàng)建一自組織特征映射newhop創(chuàng)建一Hopfield遞歸網(wǎng)絡(luò)newelm創(chuàng)建一Elman遞歸網(wǎng)絡(luò)2.網(wǎng)絡(luò)應(yīng)用函數(shù)sim仿真一個(gè)神經(jīng)網(wǎng)絡(luò)init初始化一個(gè)神經(jīng)網(wǎng)絡(luò)adapt神經(jīng)網(wǎng)絡(luò)的自適應(yīng)化train訓(xùn)練一個(gè)神經(jīng)網(wǎng)絡(luò)3.權(quán)函數(shù)dotprod權(quán)函數(shù)的點(diǎn)積ddotprod權(quán)函數(shù)點(diǎn)積的導(dǎo)數(shù)distEuclidean距離權(quán)函數(shù)normprod規(guī)范點(diǎn)積權(quán)函數(shù)negdistNegative距離權(quán)函數(shù)mandistManhattan距離權(quán)函數(shù)linkdistLink距離權(quán)函數(shù)4.網(wǎng)絡(luò)輸入函數(shù)netsum網(wǎng)絡(luò)輸入函數(shù)的求和dnetsum網(wǎng)絡(luò)輸入函數(shù)求和的導(dǎo)數(shù)5.傳遞函數(shù)

3、hardlim硬限幅傳遞函數(shù)hardlims對(duì)稱(chēng)硬限幅傳遞函數(shù)purelin線(xiàn)性傳遞函數(shù)tansig正切S型傳遞函數(shù)logsig對(duì)數(shù)S型傳遞函數(shù)dpurelin線(xiàn)性傳遞函數(shù)的導(dǎo)數(shù)dtansig正切S型傳遞函數(shù)的導(dǎo)數(shù)dlogsig對(duì)數(shù)S型傳遞函數(shù)的導(dǎo)數(shù)compet競(jìng)爭(zhēng)傳遞函數(shù)radbas徑向基傳遞函數(shù)satlins對(duì)稱(chēng)飽和線(xiàn)性傳遞函數(shù)6.初始化函數(shù)initlay層與層之間的網(wǎng)絡(luò)初始化函數(shù)initwb閾值與權(quán)值的初始化函數(shù)initzero零權(quán)/閾值的初始化函數(shù)initnwNguyen_Widrow層的初始化函數(shù)initconConscience閾值的初始化函數(shù)midpoint中點(diǎn)權(quán)值初始化函數(shù)7.

4、性能分析函數(shù)mae均值絕對(duì)誤差性能分析函數(shù)mse均方差性能分析函數(shù)msereg均方差w/reg性能分析函數(shù)dmse均方差性能分析函數(shù)的導(dǎo)數(shù)dmsereg均方差w/reg性能分析函數(shù)的導(dǎo)數(shù)8.學(xué)習(xí)函數(shù)learnp感知器學(xué)習(xí)函數(shù)learnpn標(biāo)準(zhǔn)感知器學(xué)習(xí)函數(shù)learnwhWidrow_Hoff學(xué)習(xí)規(guī)則learngdBP學(xué)習(xí)規(guī)則learngdm帶動(dòng)量項(xiàng)的BP學(xué)習(xí)規(guī)則learnkKohonen權(quán)學(xué)習(xí)函數(shù)learnconConscience閾值學(xué)習(xí)函數(shù)learnsom自組織映射權(quán)學(xué)習(xí)函數(shù)9.自適應(yīng)函數(shù)adaptwb網(wǎng)絡(luò)權(quán)與閾值的自適應(yīng)函數(shù)10.訓(xùn)練函數(shù)trainwb網(wǎng)絡(luò)權(quán)與閾值的訓(xùn)練函數(shù)traingd梯度下降的BP算法訓(xùn)練函數(shù)traingdm梯度下降w/動(dòng)量的BP算法訓(xùn)練函數(shù)traingda梯度下降w/自適應(yīng)lr的BP算法訓(xùn)練函數(shù)traingdx梯度下降w/動(dòng)量和自適應(yīng)lr的BP算法訓(xùn)練函數(shù)trainlmLevenberg_Marquardt的BP算法訓(xùn)練函數(shù)trainwbl每個(gè)訓(xùn)練周期用一個(gè)權(quán)值矢量或偏差矢量的訓(xùn)練函數(shù)11.分析函數(shù)maxlinlr線(xiàn)性學(xué)習(xí)層的最大學(xué)習(xí)率errsurf誤差曲面12.繪圖函數(shù)plotes繪制誤差曲面plotep繪制權(quán)和閾值在誤差曲面上的位置plotsom繪制自組織映射圖13.符號(hào)變換函數(shù)ind