How to add a constraint function to lsqnonlin

Hello, users.

I'm using Matlab and optimizaiton toolbox to find matrix x. (Please see figure attached.)

Matrix A is m*n matrix which has fixed values and matrix B is output.

I used fmincon to find matrix x when B has (almost) identical values. For this work "fun" is a function of minimizing standard deviation.

At that time I was able to use nonlcon for the condition that maximum x value is smaller than two times of minimum value. (xmax <= 2*xmin)

x = fmincon(fun,x0,A,b,Aeq,beq,lb,ub,nonlcon,options)

But when matrix B has different values, I know I cannot use fmincon anymore.

I'm trying to use lsqnonlin but it does not support constraint function.

x = lsqnonlin(fun,x0,lb,ub,options)

After numerical optimization, there is too big difference maximum value and minimum value. (xmax > 10000*xmin, it's not realistic value.)

Question. Can I get some matlab tip for this work? I mean I want to check if there is some user-subroutine or something other way to handle this.

Thanks in advance.

[Torsten]

So your problem is

min: ||A*x-B||

under the constraint

max(x_i) <= 2*min(x_i)

?

Best wishes

Torsten.