2
$\begingroup$

Let $f(x_1,...,x_n)$ be $C^0$ continuous function $R^n\to R$ defined on a compact domain $A\subset R^n$. Let $f$ be monotonously increasing w.r.t every argument in the domain of definition. I'm looking for a method to continue $f$ to the whole $R^n$ as $C^0$ continuous function monotone w.r.t every argument.

Particularly, I have a computational problem with the values of a function $f(x,y)$ sampled on 2D grid. The function is given in a compact domain of a grid and it is monotonous w.r.t. $x$ and $y$ in the domain. It should be continued to the whole grid, in any way, with the only requirement of being continuous and monotone w.r.t. every argument.

In a special case, when the domain is a square, I can construct an explicit formula for such continuation. So what I miss is a formula or an algorithm to continue the function to the bounding box of the domain, supporting the monotonic property above.

$\endgroup$
0
$\begingroup$

For nice domains, you can define $$ f(x_1,\dots,x_n) = \sup_{\substack{(y_1,\dots,y_n)\in A \\ y_1\le x_1,\, \dots, y_n\le x_n}} f(y_1,\dots,y_n). $$ But note that this doesn't always equal $f$ on the domain itself: for example, $any$ continuous function on the unit circle (not disk) in $\Bbb R^2$ vacuously is "increasing w.r.t every argument".

$\endgroup$
  • $\begingroup$ My domains in $R^2$ are full-dimensional, simply connected, not always convex. Thank you for the formula. I see two possible problems: (1) the selected $y$-set, i.e., intersection of $A$ and a corner with origin in $x$ can be empty for some points in the bounding box; (2) if $\sup$ is reached in an interior point of the $y$-set, the resulting function will be locally constant. I should refine that the functions must be strictly monotonous throughout. $\endgroup$ – Igor Sep 7 '16 at 17:24
  • $\begingroup$ Fair enough; even full-dimensional sets can be problematic though (for example, a crescent moon with corners at $(0,1)$ and $(1,0)$). $\endgroup$ – Greg Martin Sep 7 '16 at 18:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.