Lsqlin-Matrix badly scaled

1 次查看（过去 30 天）

Saad 2012-12-3

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/55502-lsqlin-matrix-badly-scaled

Dear all,

I would appreciate any comment or help on this. I have to solve a constrained linear squared problem with lsqlin which is defined as follows:

min sum(xi - dataA )^2 subject to sum(xi*dataB)=Constant xi...xn

             n = 3000;
            Aeq = vertcat(ones(1,n),Data B) ; 
          beq=matrix of constant (i.e. constraints)
            lb = zeros(n,1) ; 
            H = eye(n) ;

Please correct me if i am wrong but i think by using lsqlin my problem is translated this way: (i may be wrong)

            [x,fval] = lsqlin( sqrt(1/2).*H, sqrt(1/2).*dataA ,[],[], Aeq,beq,lb,[],initial guess, []);

I keep getting the message "Warning: Matrix is singular, close to singular or badly scaled." ( i have also tried to optimize the problem with fmincon but i get the same message)

I understand that the matrix is ill conditioned but I honestly don't know how to fix that. Any advice would be much appreciated. Thanks a lot