1

Example 2.16 in Boyd's book:

Cone of polynomials non-negative on $[0,1]$ can be defined as:

$K =\{c \in R^n | c_1 + c_2 t + \dots + c_n t^{n-1} \geq 0 \text{ for } t \in [0,1] \}$.

Its mentioned that $K$ is the cone of (coefficients of) degree $n-1$ that are non-negative on the interval $[0,1]$ and $K$ a proper cone. How polynomials satisfy that? Any explanation is appreciated. Thanks.

jhon_wick
  • 623
  • 1
    You haven't transcribed the definition of the set properly. Please double-check. – Michael Grant Feb 12 '17 at 17:52
  • 1
    But the thing to keep in mind is this: this is not a set of polynomials. After all, as its definition clearly states, it is a set of vectors in $\mathbb{R}^n$. The polynomial comes in only when deciding whether or not a particular vector $c$ is in the set; and in that definition, $c$ serves as the coefficients of a poynomial. – Michael Grant Feb 12 '17 at 17:53
  • thanks for the comment. Just updated the definition. However, I am still unclear about the example. – jhon_wick Feb 12 '17 at 18:08

1 Answers1

1

Hint: write the polynomial inequality associated with $c \in K$ and show that:

  • $0_n \in K$ where $0_n$ is the vector with all components equal to $0\,$;

  • $c \in K \implies \lambda c \in K$ for any real $\forall \lambda \ge 0\,$;

  • $c,c' \in K \implies \lambda c + (1-\lambda)c' \in K$ for any real $\forall \lambda \in [0,1]\,$.

dxiv
  • 77,867
  • Thanks. So the first point proves that $K$ is pointed, third one proves that $K$ is convex. I am not sure about the second one and how do we show that $K$ is solid? – jhon_wick Feb 12 '17 at 20:31
  • @jhon_wick The second one proves that it is indeed a (linear) cone. What is your definition of "solid"? – dxiv Feb 12 '17 at 20:33
  • here solid implies $K$ has nonempty interior. – jhon_wick Feb 12 '17 at 20:36
  • @jhon_wick Note for example that all vectors with non-negative components $c_k \ge 0,$ are in $K,$. – dxiv Feb 12 '17 at 20:40
  • how does the linearity show it has nonempty interior? im assuming the interior set of $K_+[\alpha,\beta]$ is the set ${ p\in\mathbb{R}[\tau] \quad | \quad det(p)\leq d, p(\tau)\geq 0 \quad \forall \tau \in [\alpha,\beta]}$, but how do i show this? i also thought of finding an open ball around an interior point, but i wasnt sure how to show that all polynomials within $\varepsilon$-distance are nonnegative on $[\alpha,\beta]$. – Britta Mar 27 '20 at 10:00
  • @Britta But a Convex Set with Empty Interior Lies in an Affine Set so it suffices to show that $K$ does not, for example the vectors of the standard basis lie in $K$. – dxiv Mar 27 '20 at 22:09