Can we describe the conic hull of a set with this formula:

\begin{equation*} \text{Conic} \space C = \{\textbf x=\textbf x_0+ \theta \textbf{v} | \textbf x_0 = \textbf 0_n , \textbf{v} \in \text{conv} \space C , \theta\in \ \mathbb R_{+}\} \end{equation*}

where $\text{conv} \space C$ is the covex hull of the set $C$.

Definition of conic hull:

$\text{Conic} \space C = \{\Sigma_{i=1}^{k}\theta_i x_i | x_i \in C, \theta_i\in\mathbb R_+\}$

it is the set of conic combinations of some points of $C$.

My proof:

$K = \{\theta \textbf v| \textbf v \in \text{conv} \space C, \theta \in \mathbb R_+ \} = \text{conic C}$

For proving this: we should show $K \subseteq \text{conic C}$ and $\text{conic C} \subseteq K$.

a. $K \subseteq \text{conic C}$:

suppose $\textbf{x} \in K$, then: $\textbf{x} = \theta \textbf{v}$ where $\textbf{v} \in \text{conv} \space C$, $\theta \in \mathbb R_+$ $\Rightarrow$ $\textbf{v} = \sum_{i} {\theta_i} x_i$, $x_i \in C$, $\Rightarrow$ $\frac{x}{\theta} = \sum_{i} {\theta_i}x_i$ $\Rightarrow$ $x = \sum_{i} {\alpha_i x_i}$ where ${\alpha_i} = \theta_i \theta$ $\in \mathbb R_+$ $\Rightarrow$ $\textbf{x} \in \text{conic C}$.

b. $\text{conic C} \subseteq K$:

suppose $\textbf{x} \in \text{conic C}$, then: $\textbf{x} = \sum_i{\theta}_i x_i$, where $x_i \in C$, $\theta_i \in \mathbb R_+$ $\Rightarrow$ $\frac{{\textbf{x}}}{\theta} = \sum_i{\frac{\theta_{i}}{\theta}}{x_i} $ where $\theta = \sum_i{\theta_i}$ $\Rightarrow$ $\frac{{\textbf{x}}}{\theta} = \sum_i{\lambda_{i}}{x_i} $ where $\sum_i {\lambda_{i}} = 1$ $\Rightarrow$ $\textbf{x} = \theta \textbf{v}$ where $\textbf{v} = \sum_i{\lambda_{i}}{x_i} \in \text{conv} \space C$.

Hence: $K = \text{conic C}$.


1 Answer 1


Let $C \in \mathbb{R}^n$ and define:

$$\text{Co}(C)=\{\theta\cdot v\in \mathbb{R}^n\,|\,v \in \text{conv}(C), \theta \in \mathbb{R}_+\}$$

$\text{$(1)$: $\text{Co}(C)\subset \text{Conic}(C)$}$

Let $x=\theta\cdot v\in \text{Co}(C)$, where $v \in \text{conv}(C)$ and $\theta \in \mathbb{R}_+$. Since $v \in \text{conv}(C)$, there must be some $n \in \mathbb{N}$ and $\alpha_i \geq0,v_i \in C$ with

$$\sum_{i=1}^n\alpha_i\cdot v_i=v\\ \sum_{i=1}^n\alpha_i=1$$

Then $x=\sum_{i=1}^n\underbrace{(\theta\cdot \alpha_i)}_{\in \mathbb{R}_+}\cdot v_i$ and hence lies in $\text{Conic}(C)$.

$\text{$(2)$: $\text{Conic}(C)\subset \text{Co}(C)$}$

Indeed, let $x=\sum_{i=1}^k\theta_i\cdot x_i \in \text{Conic}(C)$, where $x_i \in C$ and $\theta_i\in\mathbb{R}_+$. Let $\theta=\sum_{i=1}^k\theta_i$. If $\theta=0$, then all $\theta_i$ are zero and $x=0$, so $x$ lies in $\text{Co}(C)$.

Now, suppose $\theta>0$ and let $\alpha_i=\frac{\theta_i}{\theta}$. Then $\alpha_i\geq0$ and $\sum_{i=1}^k \alpha_i=1$. Then $v=\sum_{i=1}^k\alpha_i\cdot x_i$ lies in $\text{conv}(C)$. Finally, it suffices to check that $x=\theta\cdot v$ to conclude that $x \in \text{Co}(C)$.

Together, inclusions $(1)$ and $(2)$ imply that $\text{Co}(C)=\text{Conic}(C)$, so the answer is yes.

  • $\begingroup$ Dear @Fimpellizieri, my proof is also the same as you. Thanks for your final "yes". I needed to get sure. $\endgroup$
    – Amin
    Nov 30, 2016 at 6:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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