I am using the CVX tool from Boyd for convex optimization.
I need to implement the convex function [latex]\frac{1}{xy}[/latex] for positive numbers.
I figured one way which is the same method used when convexifying a Geommetric Program. However for some reason it fails to do a simple case. The code in matlab is:
cvx_begin
variables p q r
minimize( r )
p<=log(2)
q<=log(3)
0<=p+q+r
cvx_end
[latex]\exp{r}[/latex] is [latex]\frac{1}{2*3}[/latex]. However it fails and says
??? Index exceeds matrix dimensions.
Error in ==> cvxprob.eliminate at 215
temp = reshape(ndxs(temp),size(temp));
Error in ==> cvxprob.solve at 14
[ At, cones, sgn, Q, P, dualized ] = eliminate( prob, true );
Error in ==> cvx_end at 77
solve( prob );
