用LS-SVM对下面一个叠加信号做预测

hewei204114

以下是：用LS-SVM对下面一个叠加信号做预测,由其中的1024点预测随后的200点数据。但是在运行时出现了错误，谁能帮忙具体指点一下！不胜感激。
clear
clc
t=1/512:1/512:1024/512;
X=2*cos(8*pi*t)+0.5*cos(4*pi*t)+0.5*cos(pi*t);
XX=X(625:1024)';%生成模型的数据点,如果取全部数据点的话,会相当慢
plot(XX);
%以下是模仿LS-SVM说明书上的例子
delays = 20;
Xu = windowize(XX,1:delays+1);
Xtra = Xu(1:200,1:delays); Ytra = Xu(1:200,end);
Xval = Xu(201:380,1:delays); Yval = Xu(201:380,end);
%生成最优参数
[gam,sig2] = tunelssvm({Xu(:,1:delays),Xu(:,end),'f',10,002,'RBF_kernel'},[],...
'gridsearch',{},'validate',{Xtra, Ytra, Xval, Yval});
Xt=Xu(380,2:21);
%预测200点数据
nb=200; prediction = predict({Xu(:,1:delays),Xu(:,end),...
'f',gam,sig2,'RBF_kernel'},Xt,nb);

但是在运行时出现的错误如下：
??? Input argument "startvalues" is undefined.

Error in ==> gridsearch at 70
dim = size(startvalues,2);

Error in ==> tunelssvm>simanncostfun2 at 467
cost = feval(costfun,model,costargs{:});

Error in ==> tunelssvm>@(x)simanncostfun2(x,model,costfun,costargs) at 149
        [par,fval] =
        csa(rand(2,5),@(x)simanncostfun2(x,model,costfun,costargs));

Error in ==> csa at 22
e0 = feval(herrfunc,pn,varargin{:});

Error in ==> tunelssvm at 149
        [par,fval] =
        csa(rand(2,5),@(x)simanncostfun2(x,model,costfun,costargs));

Error in ==> Untitled at 17
[gam,sig2] = tunelssvm({Xu(:,1:delays),Xu(:,end),'f',10,002,'RBF_kernel'},[],
'gridsearch',{},'validate',{Xtra, Ytra, Xval, Yval});

谁能帮忙具体指点一下！不胜感激。

ChaChing

回复 1 # hewei204114 的帖子

个人水平有限, 不懂此专业! 但怎搜不到tunelssvm此函数?? (R2009a)

meiyongyuandeze

1. >>help windowize
   
2. windowize.m not found.
   
3. Use the Help browser Search tab to search the documentation, or
   
4. type "help help" for help command options, such as help for methods.
   
5. 
   
6. >> help  tunelssvm
   
7. tunelssvm.m not found.
   
8. Use the Help browser Search tab to search the documentation, or
   
9. type "help help" for help command options, such as help for methods.

复制代码

好像我的matlab2007a中也没有tunelsscm函数，也没有windowize函数！

英云

是工具箱问题，你的工具箱不完整