Canonical: https://knowlpedia.com/convex-analysis/equivalent-characterizations-of-convex-functions/

# Equivalent characterizations of convex functions

Convexity via epigraph is equivalent to Jensen and extended Jensen inequalities

**Theorem.**
Let \(f:X\to\mathbb{R}\) be a real-valued function on a real [vector space](https://knowlpedia.com/linear-algebra/vector-space/) \(X\). The following are equivalent:

1. \(f\) is [convex](https://knowlpedia.com/convex-analysis/convex-function-via-epigraph/) (i.e., \(\mathrm{epi}(f)\) is convex).
2. (**Jensen inequality**) For all \(x,y\in X\) and \(\lambda\in(0,1)\),
\[
f(\lambda x+(1-\lambda)y)\le \lambda f(x)+(1-\lambda)f(y).
\]

3. (**Extended Jensen inequality**) For all \(m\in\mathbb{N}\), all \(x_1,\dots,x_m\in X\), and all \(\lambda_i\ge 0\) with \(\sum_{i=1}^m\lambda_i=1\),
\[
f\!\left(\sum_{i=1}^m \lambda_i x_i\right)\le \sum_{i=1}^m \lambda_i f(x_i).
\]

**Context.** Item (3) says that convexity is equivalent to "subadditivity under" [convex combinations](https://knowlpedia.com/convex-analysis/convex-combination/) of finitely many points.

**Proof sketch.** (2) ⇒ (1): show any convex combination of points in the epigraph stays in the epigraph using the inequality. (1) ⇒ (2): apply convexity of the epigraph to \((x,f(x))\) and \((y,f(y))\). (2) ⇔ (3): (3) implies (2) by taking \(m=2\); (2) implies (3) by induction and the convex-combination characterization.
