Can we say that $\sup_{t}\max_Sf(t,S)=\max_S\sup_{t}f(t,S)$ for any bounded function $f$?

45 Views Asked by At

Let $f$ be a function defined over $X\times \mathcal{C}$, where $\mathcal{C}$ is a finite collection of sets, and $X=\mathbb{R}$. Furthermore, Let $f$ be bounded on $X\times \mathcal{C}$. My question is:

Can we say that $$\sup_{t}\max_S f(t,S)=\max_S\sup_t f(t,S)?$$

I think the answer should be positive as indicated by the following quick analysis that I chalked out:

$$\sup_t\max_Sf(t,S)\ge f(t,S),\ \forall t\in X, Y\in \mathcal{C}\\\implies \sup_t\max_Sf(t,S)\ge \max_S\sup_t f(t,S).$$ Similarly establishing the reverse direction, we find the desired equality.

The above method seems legit to me, but I am not sure if I am missing something, probably some subtle aspect of $\sup$, or something related to compactness of the $X$. I really don't find anything wrong, but it will be really helpful if the experts in this site can kindly verify the validity of the statement, or provide counter-examples, and relevant references for this problem.

BTW, although this seems to me a very standard and classical problem, I could not find anything directly involving the $\sup$ and $\max$, whereas its relative, Sion's minimax theorem, which involves $\sup$ and $\inf$ and is true only under certain restrictions on $f$, is a rather famous and useful tool in many parts of analysis.

Any help is appreciated. Thanks in advance.

1

There are 1 best solutions below

3
On BEST ANSWER

You can always switch the order in which supremum is taken. Of course maximum is same as supremum when it is sup over a finite set. The proof just involves verifying that each side is less than nor equal to the other.