Week 1: Convex Sets
1 Definition of convex set
A set
C
C
C is convex if the line segment between any two points in
C
C
C lies in
C
C
C, i.e., if for any
x
1
,
x
2
∈
C
x_1, x_2 \in C
x1,x2∈C and any
θ
\theta
θ with
0
≤
θ
≤
1
0 ≤ \theta ≤ 1
0≤θ≤1, we have:
θ
x
1
+
(
1
−
θ
)
x
2
∈
C
\theta x_1+(1-\theta)x_2\in C
θx1+(1−θ)x2∈C
2 Operations preserving convexity
2.1 Affine transformation (shift, scale, rotate)
C ′ = { A x + b ∣ x ∈ C } C'=\{Ax+b|x\in C\} C′={Ax+b∣x∈C} C ′ C' C′ is a convex set ⇔ \Leftrightarrow ⇔ C C C is a convex set.
2.2 Intersection
C ′ = { x ∣ x ∈ C 1 , x ∈ C 2 } C'=\{x|x\in C_1, x\in C_2\} C′={x∣x∈C1,x∈C2} C ′ C' C′ is a convex set ⇔ \Leftrightarrow ⇔ C 1 C_1 C1, C 2 C_2 C2 are convex sets.
3 Examples
3.1 Hyperplanes
{ x : a T x = b } , a ≠ 0 \{x:a^Tx=b\}, a\neq0 {x:aTx=b},a=0
3.2 Halfspaces
{ x : a T x ≤ b } , a ≠ 0 \{x:a^Tx\leq b\}, a\neq0 {x:aTx≤b},a=0
3.3 Convex hull of x 1 … x n x_1 \dots x_n x1…xn
{ x : x = ∑ i = 1 n θ i x i , θ i ≥ 0 , ∑ i = 1 n θ i = 1 } \{x:x=\sum_{i=1}^n \theta_ix_i, \theta_i\geq 0, \sum_{i=1}^n \theta_i=1\} {x:x=i=1∑nθixi,θi≥0,i=1∑nθi=1}
3.4 Conic combination of x 1 … x n x_1 \dots x_n x1…xn
{ x : x = ∑ i = 1 n θ i x i , θ i ≥ 0 } \{x:x=\sum_{i=1}^n \theta_ix_i,\theta_i\geq 0\} {x:x=i=1∑nθixi,θi≥0}
3.5 Affine combination of x 1 … x n x_1 \dots x_n x1…xn
{ x : x = ∑ i = 1 n θ i x i , θ ∈ R } \{x:x=\sum_{i=1}^n \theta_ix_i, \theta \in \mathbb{R}\} {x:x=i=1∑nθixi,θ∈R}
3.6 Ellipse, norm balls
{ x : ( x − c ) T M ( x − c ) ≤ 1 , M ≥ 0 } \{x:(x-c)^TM(x-c)\leq 1, M\geq 0 \} {x:(x−c)TM(x−c)≤1,M≥0} { x : ∣ ∣ x ∣ ∣ ≤ u } \{x:||x||\leq u\} {x:∣∣x∣∣≤u}
3.7 Polyhedra
{ x : A x ≤ b , c x = d } \{x:Ax\leq b, cx=d\} {x:Ax≤b,cx=d}
3.8 All positive semidefinite (symmetric) matrices
3.9 Level sets
f f f is a convex function ⇒ \Rightarrow ⇒ any level set { x : f ( x ) ≤ c } \{x:f(x)\leq c\} {x:f(x)≤c} is a convex set.
429
