Paper 3, Section II, H

Optimization | Part IB, 2011

(i) What does it mean to say a set CRnC \subseteq \mathbb{R}^{n} is convex?

(ii) What does it mean to say zz is an extreme point of a convex set C?C ?

Let AA be an m×nm \times n matrix, where n>mn>m. Let bb be an m×1m \times 1 vector, and let

C={xRn:Ax=b,x0}C=\left\{x \in \mathbb{R}^{n}: A x=b, x \geqslant 0\right\}

where the inequality is interpreted component-wise.

(iii) Show that CC is convex.

(iv) Let z=(z1,,zn)Tz=\left(z_{1}, \ldots, z_{n}\right)^{T} be a point in CC with the property that at least m+1m+1 indices ii are such that zi>0z_{i}>0. Show that zz is not an extreme point of CC. [Hint: If r>mr>m, then any set of rr vectors in Rm\mathbb{R}^{m} is linearly dependent.]

(v) Now suppose that every set of mm columns of AA is linearly independent. Let z=(z1,,zn)Tz=\left(z_{1}, \ldots, z_{n}\right)^{T} be a point in CC with the property that at most mm indices ii are such that zi>0z_{i}>0. Show that zz is an extreme point of CC.

Typos? Please submit corrections to this page on GitHub.