1/2*norm(beta)^2 error using .* disciplined convex programming error


(Chen Zhao) #1

can anyone help me? I a new learner to CVX and convex optimization,

here is my code

KeXi=zeros(NumberofTrainingData,1);
C=1;
tic;
cvx_begin

variable OutputWeight(NumberofHiddenNeurons,NumberofOutputNeurons)

minimize( 1/2*norm(OutputWeight)*norm(OutputWeight)+C/2*norm(KeXi)*norm(KeXi));

subject to

    Y-H*OutputWeight == KeXi;

  %  Epsilon.*ones(N,1)-norm( x, Inf ) <= 0

cvx_end
toc

the error information is

I wonder the norm of outweight is convex so the square of it is convex,
why the matlab give this information?
at first I use power_p to express the square of norm outweight, it also error, so I change it to the multiply form.

can anyone help me with the error information? Many Thanks.


(Mark L. Stone) #2

Use ` square_pos .

Please read http://cvxr.com/cvx/doc/funcref.html#new-functions