第4章神經(jīng)網(wǎng)絡(luò)工具箱函數(shù)

1、第4章神經(jīng)網(wǎng)絡(luò)工具箱函數(shù),,本章要點(diǎn)介紹神經(jīng)網(wǎng)絡(luò)工具箱中的通用函數(shù)用法介紹各類神經(jīng)網(wǎng)絡(luò)工具箱函數(shù)用法,,4.2 神經(jīng)網(wǎng)絡(luò)工具箱中的通用函數(shù)4.2.1 神經(jīng)網(wǎng)絡(luò)仿真函數(shù)神經(jīng)網(wǎng)絡(luò)仿真函數(shù)sim()功能：主要用于對(duì)神經(jīng)網(wǎng)絡(luò)進(jìn)行仿真。調(diào)用格式： [Y,Pf,Af,E,perf] = sim(net,P,Pi,Ai,T)[Y,Pf,Af,E,perf] = sim(net,{Q TS},Pi,Ai,T)[Y,Pf,Af,E,pe

2、rf] = sim(net,Q,Pi,Ai,T),,4.2.2神經(jīng)網(wǎng)絡(luò)訓(xùn)練及學(xué)習(xí)函數(shù)1. train（）函數(shù)功能：用于對(duì)神經(jīng)網(wǎng)絡(luò)進(jìn)行訓(xùn)練。調(diào)用格式： [net,tr,Y,E,Pf,Af] = train(net,P,T,Pi,Ai)2. learnp()函數(shù)功能：用于神經(jīng)網(wǎng)絡(luò)權(quán)值和閾值的學(xué)習(xí)。調(diào)用格式：[dW,LS] = learnp(W,P,Z,N,A,T,E,gW,gA,D,LP,LS) 3. learnpn

3、()函數(shù)功能：用于神經(jīng)網(wǎng)絡(luò)歸一化權(quán)值和閾值的學(xué)習(xí)。調(diào)用格式：[dW,LS] = learnpn(W,P,Z,N,A,T,E,gW,gA,D,LP,LS)info = learnpn(code),,4.2.3神經(jīng)網(wǎng)絡(luò)初始化函數(shù)1. init（）函數(shù)功能：對(duì)神經(jīng)網(wǎng)絡(luò)的參數(shù)進(jìn)行初始化。調(diào)用格式：net = init(net) 2. initlay（）函數(shù)功能：對(duì)層-層結(jié)構(gòu)神經(jīng)網(wǎng)絡(luò)的參數(shù)進(jìn)行初始化。調(diào)用格式：net =

4、 initlay(net) 3. initnw（）函數(shù)功能：對(duì)一個(gè)層進(jìn)行初始化。調(diào)用格式：net = initnw(net,i),,4. initwb（）函數(shù)功能：對(duì)一個(gè)層進(jìn)行初始化。調(diào)用格式：net = initwb(net,i) 5. revert（）函數(shù)功能：將神經(jīng)網(wǎng)絡(luò)中的權(quán)值和閾值恢復(fù)為初始值。調(diào)用格式：net = revert(net),,4.2.4神經(jīng)網(wǎng)絡(luò)輸入函數(shù)1. netsum（）函數(shù)功

5、能：對(duì)輸入求和。調(diào)用格式：N = netsum(Z1,Z2,...,Zn) 2. netprod（）函數(shù)功能：對(duì)輸入數(shù)據(jù)求和。調(diào)用格式：N = netprod(Z1,Z2,...,Zn)info = netprod(code),,4.2.5神經(jīng)網(wǎng)絡(luò)傳遞函數(shù)1. hardlim（）函數(shù)功能：硬限幅傳遞函數(shù)。調(diào)用格式：A = hardlim(N,FP) 2. hardlims（）函數(shù) 功能：對(duì)稱硬限幅傳遞函數(shù)。

6、調(diào)用格式：A = hardlims(N,FP),4.3 感知器神經(jīng)網(wǎng)絡(luò)工具箱函數(shù),4.3.1 感知器神經(jīng)網(wǎng)絡(luò)創(chuàng)建函數(shù)newp（）函數(shù)調(diào)用格式：net = newp(PR,S,TF,LF) 4.3.2 感知器神經(jīng)網(wǎng)絡(luò)顯示函數(shù)1. plotpc（）函數(shù)功能：用于在感知器向量圖中繪制分界線。調(diào)用格式：plotpc(W,B)plotpc(W,B,H) 2. plotpv（）函數(shù)功能：用于輸入向量和目標(biāo)向量繪制。調(diào)用

7、格式：plotpv(P,T)plotpv(P,T,V),,4.3.3 感知器神經(jīng)網(wǎng)絡(luò)性能函數(shù)mae（）函數(shù)功能：用于計(jì)算輸出量和目標(biāo)量之間的平均絕對(duì)誤差。調(diào)用格式：perf = mae(E,Y,X,FP)dPerf_dy = mae('dy',E,Y,X,perf,FP)dPerf_dx = mae('dx',E,Y,X,perf,FP)info = mae(code),4.4 BP神經(jīng)網(wǎng)絡(luò)工具

8、箱函數(shù),4.4.1 BP神經(jīng)網(wǎng)絡(luò)創(chuàng)建函數(shù)1. newff（）函數(shù)功能：創(chuàng)建一前饋BP網(wǎng)絡(luò)。調(diào)用格式：net = newff(P,T,[S1 S2...S(N-l)],{TF1 TF2...TFNl}, BTF,BLF,PF,IPF,OPF,DDF) 2. newcf（）函數(shù)功能：創(chuàng)建一多層前饋BP網(wǎng)絡(luò)。調(diào)用格式：net = newcf(P,T,[S1 S2...S(N-l)],{TF1 TF2...TFN}, BTF,

9、BLF,PF,IPF,OPF,DDF) 3. newfftd（）函數(shù)功能：創(chuàng)建一前饋輸入延遲BP網(wǎng)絡(luò)。調(diào)用格式：net = newfftd(P,T,ID,[S1 S2...S(N-l)],{TF1 TF2...TFNl}, BTF,BLF,PF,IPF,OPF,DDF),,4.4.2 BP神經(jīng)網(wǎng)絡(luò)傳遞函數(shù)1. purelin（）函數(shù)功能：線性傳遞函數(shù)。調(diào)用格式：A = purelin(N,FP) 2. tansig

10、（）函數(shù)功能：線性傳遞函數(shù)。調(diào)用格式：A = tansig(N,FP) 3. logsig（）函數(shù)功能：對(duì)數(shù)S型傳遞函數(shù)。調(diào)用格式：A = logsig(N,FP),,4.4.3 BP神經(jīng)網(wǎng)絡(luò)學(xué)習(xí)函數(shù)1. learngd（）函數(shù)功能：梯度下降學(xué)習(xí)函數(shù)。調(diào)用格式：[dW,LS] = learngd(W,P,Z,N,A,T,E,gW,gA,D,LP,LS)[db,LS] = learngd(b,ones(1,Q),Z

11、,N,A,T,E,gW,gA,D,LP,LS) 2. learngdm（）函數(shù)功能：梯度下降動(dòng)量學(xué)習(xí)函數(shù)。調(diào)用格式：[dW,LS] = learngdm(W,P,Z,N,A,T,E,gW,gA,D,LP,LS)[db,LS] = learngdm(b,ones(1,Q),Z,N,A,T,E,gW,gA,D,LP,LS),4.4.4 BP神經(jīng)網(wǎng)絡(luò)訓(xùn)練函數(shù)1. trainbfg（）函數(shù)功能：準(zhǔn)牛頓BP算法函數(shù)。調(diào)用格式：[

12、net,TR] = trainbfg(net,TR,trainV,valV,testV) 2. traingd（）函數(shù)功能：梯度下降的BP算法訓(xùn)練函數(shù)。調(diào)用格式：[net,TR] = traingd(net,TR,trainV,valV,testV) 3. traingdm（）函數(shù)功能：梯度下降動(dòng)量的BP算法訓(xùn)練函。調(diào)用格式：[net,TR] = traingdm(net,TR,trainV,valV,testV),

13、4.4.5 BP神經(jīng)網(wǎng)絡(luò)性能函數(shù)1. mse（）函數(shù)功能：計(jì)算均方誤差。調(diào)用格式：perf = mse(E,Y,X,FP)dPerf_dy = mse('dy',E,Y,X,perf,FP)dPerf_dx = mse('dx',E,Y,X,perf,FP) 2. msereg（）函數(shù)功能：加權(quán)均方誤差函數(shù)。調(diào)用格式：perf = msereg(E,Y,X,FP)dPerf_dy = m

14、sereg('dy',E,Y,X,perf,FP)dPerf_dx = msereg('dx',E,Y,X,perf,FP),4.4.6 BP神經(jīng)網(wǎng)絡(luò)顯示函數(shù)1. plotperf（）函數(shù)功能：用于繪制網(wǎng)絡(luò)性能。調(diào)用格式：plotperf(TR,goal,name,epoch) 2. plotes（）函數(shù)功能：用于繪制單獨(dú)神經(jīng)元的誤差曲面。調(diào)用格式：plotes(WV,BV,ES,V)

15、 3. plotep（）函數(shù)功能：用于繪制權(quán)值和閾值在誤差曲面上的位置。調(diào)用格式：h= plotep(W,B,E)h= plotep(W,B,E,H) 4. errsurf（）函數(shù)功能：用于計(jì)算單個(gè)神經(jīng)元的誤差曲面。調(diào)用格式：errsurf(P,T,WV,BV,F),4.5.1 線性神經(jīng)網(wǎng)絡(luò)創(chuàng)建函數(shù)和設(shè)計(jì)函數(shù)1. newlin（）函數(shù)功能：用于創(chuàng)建一線性層。調(diào)用格式：net = newlin(P,S,ID,

16、LR) newlind（）函數(shù)功能：用于設(shè)計(jì)一線性層。調(diào)用格式：net = newlind(P,T,Pi),4.5 線性神經(jīng)網(wǎng)絡(luò)工具箱函數(shù),4.5.2 學(xué)習(xí)函數(shù)1. learnwh（）函數(shù)功能：Widrow_Hoff學(xué)習(xí)規(guī)則，實(shí)現(xiàn)輸出誤差的平方和最小功能。調(diào)用格式：[dW,LS] = learnwh(W,P,Z,N,A,T,E,gW,gA,D,LP,LS)[db,LS] = learnwh(b,ones(1,Q),Z,

17、N,A,T,E,gW,gA,D,LP,LS) 2. maxlinlr（）函數(shù)功能：計(jì)算線性層的最大學(xué)習(xí)速率。調(diào)用格式：lr = maxlinlr(P)lr = maxlinlr(P,'bias'),4.6.1 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)創(chuàng)建函數(shù)1. newc（）函數(shù)功能：創(chuàng)建競(jìng)爭(zhēng)層。調(diào)用格式：net = newc(PR,S,KLR,CLR) newsom（）函數(shù)功能：創(chuàng)建一自組織特征映射。調(diào)用格式：

18、net = newsom(P,[D1,D2,...],TFCN,DFCN,STEPS,IN) newlvq（）函數(shù)功能：創(chuàng)建學(xué)習(xí)向量化LVQ網(wǎng)絡(luò)。調(diào)用格式：net = newlvq(PR,S1,PC,LR,LF),4.6 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)工具箱函數(shù),4.6.2 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)傳遞函數(shù)１.　compet（）函數(shù)功能：競(jìng)爭(zhēng)性傳遞函數(shù)。調(diào)用格式：A = compet(N,FP) 2.　softmax（）函數(shù)功能：軟最

19、大傳遞函數(shù)。調(diào)用格式：A = softmax(N,FP)4.6.3 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)距離函數(shù)１.　boxdist（）函數(shù)功能：box距離函數(shù)。調(diào)用格式：d = boxdist(pos),2.　 dist（）函數(shù) 功能：歐氏距離權(quán)函數(shù)。調(diào)用格式：Z = dist(W,P,FP)D = dist(pos)3.　linkdist（）函數(shù)功能：鏈接距離函數(shù)。調(diào)用格式：d = linkdist(pos) 4. man

20、dist（）函數(shù) 功能：Manhattan距離權(quán)函數(shù)。調(diào)用格式：Z = mandist(W,P)D = mandist(pos),4.6.4 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)學(xué)習(xí)函數(shù)1. 　learnk（）函數(shù)功能：Kohonen權(quán)值學(xué)習(xí)函數(shù)。調(diào)用格式：[dW,LS] = learnk(W,P,Z,N,A,T,E,gW,gA,D,LP,LS) 2. 　learnsom（）函數(shù)功能：自組織映射權(quán)值學(xué)習(xí)函數(shù)。調(diào)用格式：[dW,LS] = l

21、earnsom(W,P,Z,N,A,T,E,gW,gA,D,LP,LS) 3. 　learnis（）函數(shù) 功能：instar權(quán)值學(xué)習(xí)函數(shù)。調(diào)用格式：[dW,LS] = learnis(W,P,Z,N,A,T,E,gW,gA,D,LP,LS),4.　 learnos（）函數(shù)功能：outstar權(quán)值學(xué)習(xí)函數(shù)。調(diào)用格式： [dW,LS] = learnos(W,P,Z,N,A,T,E,gW,gA,D,LP,LS)4.6.5 自組織競(jìng)

22、爭(zhēng)神經(jīng)網(wǎng)絡(luò)初始化函數(shù)midpoint（）函數(shù)功能：將權(quán)值初始化為輸入向量值域中心的函數(shù)。調(diào)用格式： W = midpoint(S,PR)4.6.6 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)權(quán)值函數(shù)negdist（）函數(shù)功能：負(fù)距離權(quán)值函數(shù)。調(diào)用格式： Z = negdist(W,P,FP),4.6.7 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)顯示函數(shù)poltsom（）函數(shù)功能：繪制自組織特征映射的函數(shù)。調(diào)用格式： plotsom(pos)plotsom(W,

23、D,ND)4.6.8 自組織競(jìng)爭(zhēng)神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)函數(shù)1. 　hextop（）函數(shù)功能：實(shí)現(xiàn)六角層拓?fù)浜瘮?shù)。調(diào)用格式：pos = hextop(dim1,dim2,...,dimN) 2. 　gridtop（）函數(shù) 功能：實(shí)現(xiàn)網(wǎng)格層拓?fù)浜瘮?shù)。調(diào)用格式：pos = gridtop(dim1,dim2,...,dimN),4.7.1 徑向基神經(jīng)網(wǎng)創(chuàng)建函數(shù)1. 　newrb（）函數(shù)功能：創(chuàng)建徑向基網(wǎng)絡(luò)。調(diào)用格式： [net,tr]

24、 = newrb(P,T,goal,spread,MN,DF)2. 　newrbe（）函數(shù)功能：設(shè)計(jì)準(zhǔn)確的徑向基網(wǎng)絡(luò)。調(diào)用格式：net = newrbe(P,T,spread) 3. newpnn（）函數(shù)功能：創(chuàng)建概率徑向基網(wǎng)絡(luò)。調(diào)用格式：net = newpnn(P,T,spread) 4. newgrnn（）函數(shù)功能：設(shè)計(jì)廣義回歸徑向基網(wǎng)絡(luò)。調(diào)用格式：net = newgrnn(P,T,spread),4.7 徑向

25、基神經(jīng)網(wǎng)絡(luò)工具箱函數(shù),4.7.2 徑向基神經(jīng)網(wǎng)轉(zhuǎn)換函數(shù)1. 　ind2vec（）函數(shù)功能：將數(shù)據(jù)索引轉(zhuǎn)換為向量組。調(diào)用格式：vec = ind2vec(ind) 2. 　vec2ind（）函數(shù)功能：將向量組轉(zhuǎn)換為數(shù)據(jù)索引。調(diào)用格式：ind = vec2ind(vec) 4.7.3 徑向基神經(jīng)網(wǎng)傳遞函數(shù)radbas（）函數(shù)功能：將數(shù)據(jù)索引轉(zhuǎn)換為向量組。調(diào)用格式： A = radbas(N,FP)info = radb

26、as(code),4.8.1 Hopfield網(wǎng)絡(luò)的工具箱函數(shù)1.　newhop（）函數(shù)功能：用于設(shè)計(jì)一個(gè)Hopfield反饋網(wǎng)絡(luò)調(diào)用格式：net=newhopnet=newhop（T）2.　satlins（）函數(shù)功能：用于Hopfield網(wǎng)絡(luò)的飽和線性傳遞函數(shù)。調(diào)用格式：A=satlins（N）4.8.2 Elman網(wǎng)絡(luò)的工具箱函數(shù)newelm（）函數(shù)功能：用于設(shè)計(jì)一個(gè)Elman反饋網(wǎng)絡(luò)。調(diào)用格式：net=