← back

Jensen's Inequality

updated 2026-09-10

Remember that convex functions are such that line between two points lies above the function in 2-D

\[f(\lambda x_1+(1-\lambda)x_2) \leq \lambda f(x_1)+(1-\lambda)f(x_2)\]

This 2-D case definition is generalized to any high-dimension. Concave functions are the opposite.

Jensen’s inequality for any convex function states that:

\[f(\mathbb E[u])\leq \mathbb E[f(u)]\]

(The proof is based on expanding expectation for continuous/discrete expectations, then applying definition of convexity/concavity! Think about it!

Note that equivalently Jensen’s states that for any concave function

\[f(\lambda x_1+(1-\lambda)x_2) \geq \lambda f(x_1)+(1-\lambda)f(x_2)\]
\[f(\mathbb E[u])\geq \mathbb E[f(u)]\]

the sign just flips!