In order to determine or establish convexity of sets, it is useful to review some operators that preserve the convexity.

# Intersection

Firstly, convexity is preserved under intersection. Namely, if $C_1$ and $C_2$ are convex then $C_1\cap C_2$ is convex.

A set $C$ is called a **convex cone** if $C$ is a cone and $C$ is convex, i.e. for any $x_1,x_2\in C$ and $\lambda_1, \lambda_2 \geqslant0$ we have

# Conic combination

A **conic combination** of the points $x_1,\ldots, x_n$ is a point of form $\lambda_1x_1+\ldots+\lambda_nx_n$ with $\lambda_i\geqslant0\ \ \forall i=1,\ldots,n$.

The following table shows the difference between affine combination, convex combination and conic combination

Affine combination | Convex combination | Conic combination |
---|---|---|

Form: $\ \lambda_1x_1 +\ldots+ \lambda_nx_n$ | Form: $\ \lambda_1x_1 +\ldots+ \lambda_nx_n$ | Form: $\ \lambda_1x_1 +\ldots+ \lambda_nx_n$ |

Where $\ \lambda_i\in\mathbb R$ and $\sum_{i=1}^n \lambda_i=1$ | Where $\ \lambda_i\geqslant0 $ and $\sum_{i=1}^n \lambda_i=1$ | Where $\ \lambda_i\geqslant0$ |

# Conic hull

The set of all conic combination of points in $C$ is called the **conic hull** of $C$

**Tip:** In the definition of conic hull, we only need $\lambda_i\geqslant0$