Homogeneous functions
A function
is homogeneous of degree
if
. I’ve seen this definition before, in the proof of some inequality about integrals (a really famous one that I can’t recall the name of), but since I didn’t understand the proof, never really got interested in homogeneity until today.
I came across it again in a problem in a book: if
is homogeneous of degree
, then
. I still haven’t figured out how to prove this, but it caused me to examine the idea of homogeneity. Like, does this problem assume that
is continuous, or is that a result of the fact it’s homogeneous?
First I tried looking for individual examples of homogeneous functions, to get an overview of what they look like, and I came up with some:
,
, etc. But then I realized from the first example, that a metric has to be homogeneous of degree 1, by definition— or I thought I did, until I realized this holds true only if
.
Then I tried generalizing my thought process
to higher dim spaces, and realized that because of homogeneity,
has only to be defined on the unit ball, and then for any
,
for some
in the unit ball.
This gives us a way to construct arbitrary homogeneous functions, given just their values on the unit ball. Of course, you need to be careful that for each
,
. So, if
and
odd (even), then
has to be odd (even) to be homogenous of degree
, for example.
I realized from this ability to construct arbitrary homogeneous functions that homogeneity does not imply continuity, so the problem that got me started isn’t well-stated.
All of these observations can be made directly from the definition of homogeneity, but I think the thought process that lead to them in this particular instance was pretty neat.
Possibly relevant posts:
- NSF Personal Statement (11/3/2005)
- Weierstrass’ criterion (8/13/2007)
- Calculus of Variations (11/11/2003)
Did you prove the original problem? ie. say f:R^n -> R and f(tx) = t^n f(x). You want to prove in
general that
= nf(x)
Hint: Differentiate the homogentity condition with respect to ‘t’ using the chain rule.
denotes the standard inner product in R^n.
Comment by Marc — 3/23/2005 @ 8:22 am