Is there a unique saddle value for a convex/concave optimization?

1k Views Asked by At

Here is the question:

Consider a function of two variables $f(x,y)$ on some compact domain. Let it be convex in $x$ and concave in $y$. Is it true that $f(x,y)$ has a uniqe saddle point (stationary point)? Is $(\hat x,\hat y)=\arg \min_x\max_y f(x,y)$, the argument of saddle point, unique? Is there any role of being strictly concave (or convex) in this case?

My guess is that $f(x,y)$ owns a unique saddle point because of the convexity and concavity of the function. I also guess that if the function is strictly concave-convex, then, the argument of the saddle point is also unique.

Any reference for a proof will also suffice. Thx.