Hi everyone, I have an optimization problem of the form,
((x_1+x_2 ))/((2^(x_1/B)+2^(x_2/A)+y) ).
A and B are known constants. The variables are x_1, x_2 and y. All my inequality constraint sets are convex and equality constraint sets are affine. I strongly believe this is Quasiconvex since the numerator is affine and the denominator is a convex function.
In solving it through a sequence of convex feasibility problems, I introduce a new and additional constraint of the form,
where η is known (already solved for). Because of this new constraint I understand I cannot use CVX since
is a concave function and not convex. Now my question; please is there anyway that I can express this additional constraint in a convex form??
|The objective function can be rewritten as
,which can be solved by firstly finding by the bisection method and then solving the feasibility program for the variables a and b.
I solved the same problem, but with the inverse of the objective function and the constraint for the feasibility program as shown below using CVX.
And for the two cases, I had the same optimal solutions for the variables a and b.
I am not sure if what I have done above is right (solving the inverse). But if it is right, then for my optimization problem, I can take the inverse (which will be ratio of a convex function and an affine function (also positive concave) ) and write the additional constraint in a suitable form to fit the new objective function.