• ↑↓ pour naviguer
  • pour ouvrir
  • pour sélectionner
  • ⌘ ⌥ ↵ pour ouvrir dans un panneau
  • esc pour rejeter
⌘ '
raccourcis clavier

a convex set is essentially a set that intersects every line in a line segment 1

ELI5: a convex function graph is shaped like a cup \cup, where as a concave function graph is shaped like a cap \cap

formal definition

Let XX be a convex subset of a real vector space and let f:XRf: X \to \mathbb{R} be a function.

Then ff is called convex iff any of the following equivalent holds:

  1. 0t1x1,x2Xf(tx1+(1t)x2)tf(x1)+(1t)f(x2)\forall 0 \le t \le 1 \cap x_{1}, x_{2} \in X \mid f(tx_{1}+(1-t)x_{2}) \le tf(x_{1}) + (1-t)f(x_{2})
  2. 0t1x1,x2X where x1x2f(tx1+(1t)x2)tf(x1)+(1t)f(x2)\forall 0 \le t \le 1 \cap x_{1}, x_{2} \in X \text{ where } x_{1} \neq x_{2} \mid f(tx_{1}+(1-t)x_{2}) \le tf(x_{1}) + (1-t)f(x_{2})

This is also known as the Jensen’s inequality:

f(tx1+(1t)x2)tf(x1)+(1t)f(x2)f(tx_{1} + (1-t)x_{2}) \le tf(x_{1}) + (1-t)f(x_{2})

probability theory

The form as follows:

Given XX is a random variable with φ\varphi a convex function, then

φ(E[X])E[φ(X)]\varphi(E[X]) \le E[\varphi{(X)}]