I have checked the convexity of this function
for 0<x<1, 0<y<1but I cannot write it using DCP rules. Any suggestion on how to express it using DCP rules?
I also checked the concavity, and found it to be neither convex nor concave, even when restricted to 0 < x < 1, 0 < y < 1.
For instance, at x=0.8, y=0.2, the Hessian has eigenvalues of -0.16 and 246.06. Therefore, it can’t be used in CVX. Your only hope of using it in CVX is if the context of its use allows a reformulation without explicitly forming the function; but I doubt that is the case.
1 Like
Thanks @Mark_L_Stone ! Why do you think the plot shows convexity (I have generated that from wolfram alpha) for x,y in that interval but the Hessian evaluated at that point is not PSD?
I don’t know. Determining concavity from 3D plots is outside my expertise. it is also not an approved method per the link I provided.
It actually is convex for a large portion of 0 < x < 1, 0 < y < 1. I am not aware of any point in that region at which it is concave,
1 Like
The plot is not very precise in the area you need. Plug in Mark’s y=0.2 and plot the function of one variable x, then the problem is clear.
1 Like
How does it look? @Michal_Adamaszek
