L.

Vandenberghe

EE236B Winter 2016

Additional problems for homework #2

1. Partial addition and intersection. The intersection and sum

C1 C2 = {x | x C1 , x C2 },

C1 + C2 = {y1 + y2 | y1 C1 , y2 C2 }

of two convex sets C1 and C2 are convex sets. The following is an extension of these
two operations. Suppose C1 and C2 are convex sets in Rm+n . Define
C = {(x, y1 + y2 ) | x Rm , y1 , y2 Rn , (x, y1 ) C1 , (x, y2 ) C2 }.
Show that C is convex.
2. Conic representation of convex set. Let C be a nonempty convex set in Rn . Define
K = {(x, t) Rn+1 | t 0, x tC}.
(For a scalar t, the set tC is defined as tC = {ty | y C}.) Show that K is a convex
cone.
An example of this construction, with C a unit norm ball, is the norm cone C =
{(x, t) | t 0, kxk t} of the lecture notes, page 2-8.
3. Polar of a convex set. The polar of a set C Rn is defined as
C = {y Rn | y T x 1 for all x C}.
(a) Show that C is convex (even if C is not).
(b) Suppose C is a cone. How are the polar C and the dual cone C related?
(c) What are the polars of the following sets?
C1 = {x | kxk2 1},

C2 = {x | kxk1 1},

C3 = {x | 1T x = 1, x  0}.

(Here 1 denotes the vector of ones.)

(d) By definition of C , we have
y T x 1 for all x C, y C .
(For the set C1 this is the Cauchy-Schwarz inequality.) Show that this implies
that C (C ) .
1

(e) Suppose C is closed, convex, and 0 int C. Show that

(C ) = C.
Given the result of part (d), it is sufficient to show that (C ) C. In other
words, show that if x 6 C then x 6 (C ) .
Hints. Use the strict separating hyperplane theorem of page 49 of the textbook.
If C is a closed convex set and x 6 C then there exists a vector a 6= 0 and a
scalar b such that
aT x > b,
aT z b for all z C.
Then use the fact that 0 int C to argue that one can take b = 1. (The condition
0 int C means that there exists a small ball B = {u | kuk2 } with  > 0 and
B C.)