请教有关逆迭代法的程序

jianping

<P>K=0.25*[5,3;3,5];<BR>M=[9,7;7,9]./16;<BR>q=2;<BR><BR>function [freque,eigvec,XN0H,alafa]=nidiedaifa(K,M,q)</P>
<P>alafa=zeros(q-1,1);<BR>freque=zeros(q,1);<BR>eigvec=zeros(q,q);</P>
<P>XN0H=zeros(q,1);</P>
<P>%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% initial vector ， important!<BR>XN0=ones(q,1);<BR>%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%<BR>  <BR>  <BR>  for i=1:1:q<BR>  if i==1<BR>  XN1=XN0;<BR>else<BR>  j=i-1;<BR>  alafa(j)=eigvec(:,j)'*M*XN0;<BR>  XN0H=XN0;<BR>  for m=1:1:j<BR>  XN0H=XN0H-alafa(m)*eigvec(:,m);              %%%%%%%%%%%%%%%%%%%%%%%%%  Gram-Schmidt 正交化<BR> end  <BR>  XN1=XN0H;<BR>end<BR>  <BR>XNH=inv(K)*M*XN1;<BR>XN=XNH./(sqrt(XNH'*M*XNH));<BR>LAM0=(XNH'*K*XNH)./(XNH'*M*XNH);</P>
<P>rate=0.1;</P>
<P>while (rate&gt;1.0e-6)<BR>    XN1=XN;<BR>    XNH=inv(K)*M*XN1;<BR>    XN=XNH./(sqrt(XNH'*M*XNH));<BR>    LAM=(XNH'*K*XNH)./(XNH'*M*XNH);<BR>    rate=abs((LAM-LAM0)/LAM);<BR>    LAM0=LAM;<BR>end  <BR>freque(i)=LAM;<BR>eigvec(:,i)=XN;<BR>end<BR><BR><BR>%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%<BR>上面的逆迭代法求特征值的程序,为什么总收敛于一个最小的特征值,我已经进行了Schmidt正交化了啊.如果我把初始迭代向量XN0=ones(q,1);改成XN0=zeros(q,1);XN0(q-1,1)=1;,又可以收敛到不同的特征值了.这到底是什么原因?哪位高人能告诉我初始迭代向量的选取方法吗?使结果收敛于不同的特征值和特征向量.</P>