Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Assume that $I$ is a countable set, and we have $u_i\in \mathbb{R}^n$ for $i\in I$. Suppose that $v=\sum_{i\in I} a_i u_i$ and $\sum_{i\in I}a_i=1$ and $a_i\geq 0$.

Can one show that there exists a finite $J\subseteq I$ and $b_j$ for $j\in J$ such that $\sum_{j\in J}b_j=1$, $b_j\geq 0$, and $v=\sum_{j\in J} b_j u_j$?

Is this some well-known result?

share|cite|improve this question
Carathéodory's theorem almost says that you can get away with just $n+1$ elements in $J$. Except here you start with an infinite convex combination, and then the theorem does not apply, at least not without further work. – Harald Hanche-Olsen Oct 4 '12 at 10:41
I second joriki's comment. Do you not know about accepting answers to questions you post on this site? – Gerry Myerson Oct 4 '12 at 12:34
up vote 5 down vote accepted

Yes, one can.

Without loss of generality, assume $a_i\gt0$ and let $I=\mathbb N$. For any subset $A\subseteq\mathbb N$, call

$$ W_A=\frac{\sum_{i\in A}a_iu_i}{\sum_{i\in A}a_i} $$

the weighted average for $A$. If the $u_i$ all lie in some proper affine subspace of $\mathbb R^n$, we can restrict to that subspace, so without loss of generality we can assume that their affine span is $\mathbb R^n$. Then at some finite index $k$ there are positive contributions from $n+1$ affinely independent vectors $A=\{u_{i_1},\dotsc,u_{i_{n+1}}\}$, and $W_A$ is therefore in the interior of their convex hull $H$. Thus, since both $\sum_ia_i$ and $\sum_ia_iu_i$ converge (so their tails converge to zero) there is some $l\ge k$ such that the weighted average remains inside $H$ if the tail $\{i\in\mathbb N\mid i\ge l\}$ is added to $A$. Then the desired finite sum can be obtained by replacing the contributions from the tail by corresponding linear combinations of $u_{i_1},\dotsc,u_{i_{n+1}}$.

share|cite|improve this answer
I don't understand the argument. Let $A^\prime=\{i_1,\dots, i_{n+1}\} \cup\{i\in\mathbb{N}: i\geq l\}$. How do we know that $v=W_{A^\prime}$? – Byron Schmuland Jul 24 '14 at 0:21

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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