Posts

Showing posts from 2020

Symmetric functions and their extreme points

A symmetric function $f(x_1, \dots, x_n)$ is one that is invariant to permutations of $x$.  In the 2D case, this means $f(a,b) = f(b,a)$.  One basic example of a symmetric function is the area of a rectangle as a function of its side lengths, where $f(a,b) = a b$.  It is well known that to maximize the area of a rectangle under a constraint on the perimeter, you should set all side lengths equal, to form a square.  In this case, the maximum of this symmetric function occurs when $a=b$.

While in general it is not always true that the extreme points of a symmetric function occurs when $a=b$, there are special cases when it does hold.  For example, if $f(a,b)$ is a convex (concave) function, then a global minimum (maximum) will occur when $a=b$.  Of course, the function above is not concave, but this property still holds.  So the question arises: in what other situations does this nice property hold?

In this blog post, I will state a general result that I found today regarding this probl…