This idea can be generalized to more than two points.
where θ1+⋯+θk=1
We refer to a point can be expressed as the following form as an affine combination of the points x1,…xk
affine set : contains every affine combination of its points
If C is an affine set and x0∈C, then the set
is a subspace, i.e., closed under vector addition and scalar multiplication.
Proof
Let v1,v2∈V,α,β∈R. Then v1+x0,v2+x0∈C
Moreover,
That means αv1+βv2∈V
Therefore V is closed under vector addition and scalar multiplication.
Thus, the affine set C can be expressed as
The set of all affine combinations of points in some set C⊂Rn is called the affine hull of C, and denoted aff C
Affine dimension and relative interior
We define the dimension of an affine set C as the dimension of the subspace V. But it is not always consistent with other definitions of dimension.
Let’s think about the unit circle in R2. Its affine dimension is two. By most definitions of dimension, however, the unit circle in R2 has dimension one.
We define the relative interior of the set C, denoted relint C, as its interior relative to aff C
Convex set
line segment : all points betweenx1 and x2
where 0≤θ≤1
convex set : contains line segment between any two points in the set
where 0≤θ≤1
List of convex set
convex hull
convex cone/ norm cone
half-space
euclidean ball / norm ball
ellipsoid
polyhedra
simplex
Convex combination and convex hull
convex combination : any point x of the form
where θ1+⋯+θk=1,θi≥0
convex hull : set of all convex combinations
The convex hull of a set C, denoted convC, is the set of all convex combinations of points in C
Convex cone
A set C is called a cone if for every x∈C and θ≥0, we have θx∈C.
A set C is a convex cone if it is convex and cone, which means that
A point of the form
where θ1,…,θk≥0
is called a conic combination of x1,…,xk.
Hyperplane
A hyperplane is a set of the form
where a∈Rn,a=0,b∈R
Let
Then,
where x0∈{x∣aTx=b}
Halfspace
A hyperplane divides Rn into two halfspaces. A (closed) halfspace is a set of the form
where a=0
Euclidean balls and ellipsoids
A Euclidean ball in Rn has the form
where r>0
We can expressed above term differently
A Euclidean ball is a convex set
Proof
Let x1,x2∈Br(xc),θ≥0
So, θx1+(1−θ)x2∈Br(xc).
Therefore, a euclidean ball is convex set.
A ellipsoid in Rn has the form
where P=PT≻0 (i.e. P is symmetric and positive definite)
We can expressed above term differently
where A is square and nonsingular.
Norm balls and norm cones
A norm ball is the generalization of a Euclidean ball
A norm cone associated with the norm ∥⋅∥ is the set
As we proved in Euclidean ball case, a norm ball is also a convex set. Moreover, a norm cone is a convex set
Proof
Let (x1,t1),(x2,t2)∈C,θ≥0, i.e.
We want to show that
It is equivalent to check whether
Then, θx1+(1−θ)x2∈C
Therefore, C is a convex set.
Polyhedra
A polyhedron is defined as the solution set of a finite number of linear equalities and inequalities
It will be convenient to use the compact notation
where the symbol ⪯ denotes vector inequality or component-wise inequality in Rn.
A polyhedron is thus the intersection of a finite number of half-space and hyperplanes.
A polyhedron is a convex set
Proof
Since half space and hyperplane are convex set, it is enough to show that intersection of convexs set is also a convex set.
Let A,B is a convex set and x1,x2∈A∩B,θ≥0
Since x1,x2∈A
For the similar reason,
So, θx1+(1−θ)x2∈A∩B
Therefore, the intersection of a convex sets is also a convex set.
By using this conclusion, we can easily show that a polyhedron is a convex set.
Simplex
Suppose the k+1 points v0,…,vk∈Rn are affinely independent, which means v1−v0,⋯,vk−v0 are linearly independent
The simplex determined by them is given by
The affine dimension of this simplex is k, so it is sometimes referred to as a k-dimensional simplex in Rn
The simplex is so-named because it represents the simplest possible polytope in any given dimension. For example,
a 0-dimensional simplex is a point
a 1-dimensional simplex is a line segment
a 2-dimensional simplex is a triangle
a 3-dimensional simplex is a tetrahedron
Specifically, a k-simplex is a k-dimensional polytope that is the convex hull of its k + 1 vertices.
To describe the simplex as a polyhedron, we proceed as follows.
By definition, x∈C iff x=θ0v0+⋯+θkvk for some θ⪰0 with 1Tθ=1. Define y=(θ1,…,θk) and
we can say that x∈C iff
Since the columns in B are linearly independent, the rank of B is k. Therefore, there exists a nonsingular matrix A=[A1A2]∈Rn×n such that
Therefore,
From this we can see that x∈C if and only if
In other words, we have x∈C iff
which is a set of linear equalities and inequalities in x. Therefore, we can express a simplex as a polyhedron.
Positive semidefinite cone
We use the notation Sn to denote the set of symmetric n×n matrices
Note that Sn is a convex set.
Proof
Let X,Y∈Sn,θ≥0
θX+(1−θ)Y is also symmetric.
So, θX+(1−θ)Y∈Sn
Therefore, Sn is a convex set.
We use the notation S+n to denote the set of symmetric positive semi-definite matrices
In mathematics, these are equivalent
Note that S+n is a convex cone.
Proof
Let X,Y∈S+n, α,β≥0
For given z,
By using the properties of inner-product,
That means αX+βY∈S+n
Therefore, S+n is a convex cone.
The notation S++n to denote the set of symmetric positive definite matrices
Operations that preserve convexity
List of preserving convexity
Intersection
Affine functions
scaling and translation
projection
sum of two sets
partial sum
Cartesian product
Perspective function
Linear fractional function
Intersection
Convexity is preserved under intersection: if S1 and S2 are convex, then S1∩S2 is convex.
Proof
Let A,B is a convex set and x1,x2∈A∩B,θ≥0
Since x1,x2∈A
For the similar reason,
So, θx1+(1−θ)x2∈A∩B
Therefore, the intersection of a convex sets is also a convex set.
This property extends to the intersection of an infinite number of sets. If Sα is convex for every α∈A, then ∩α∈ASα is convex.
Affine functions
A function f:Rn→Rm is affine if it is a sum of a linear function and a constant, i.e. if it has the form
where A∈Rm×n,b∈Rm
Suppose S⊂Rn is convex and f:Rn→Rm is an affine function. Then the image of S under f,
is convex.
Proof
Let y1,y2∈f(S), θ≥0. Then there exists x1,x2 such that
Since S is a convex set, θx1+(1−θ)x2∈S. So, θy1+(1−θ)y2∈f(S).
Therefore, f(S) is a convex set.
Similarly, if f:Rk→Rn is an affine function, the inverse image of S under f,
is convex.
Proof
Let x1,x2∈f−1(S),θ≥0. Then there exists y1,y2∈S such that
So, θx1+(1−θ)x2∈f−1(S).
Therefore, f−1(S) is a convex set.
Example : scaling and translation
If S⊂Rn is convex, α∈R and a∈Rn, then the sets αS,S+a are convex, where
Proof
If α=0, αS and S+a are trivially convex sets. Therefore, without loss of generality, we may assume α>0.
Let x,y∈αS,θ≥0. Then, x/α,y/α∈S
Since S is a convex set,
So, α(θx/α+(1−θ)y/α)=θx+(1−θ)y∈αS.
Therefore, αS is a convex set.
Similarly, x,y∈S+a,θ≥0. Then x−a,y−a∈S
Since S is a convex set
So, θ(x−a)+(1−θ)(y−a)+a=θx+(1−θ)y∈S+a
Therefore, S+a is a convex set.
Example : projection
The projection of a convex set onto some of its coordinates is convex. If S⊂Rm×Rn is convex, then
is convex.
Proof
Let x1,x2∈T,θ≥0. Then there exists y1,y2∈Rn such that
Since, S is a convex set
So, θx1+(1−θ)x2∈T.
Therefore, T is a convex set.
Example : sum of two sets
The sum of two sets is defined as
If S1 and S2 are convex, then S1+S2 is convex.
Proof
Let z1,z2∈S1+S2,θ≥0. Then there exists x1,x2∈S1,y1,y2∈S2 such that
Since θx1+(1−θ)x2∈S1,θy1+(1−θ)y2∈S2, θz1+(1−θ)z2∈S1+S2.
Therefore, S1+S2 is a convex set.
Example : partial sum
The partial sum of S1,S2∈Rn×Rm, defined as
where x∈Rn,yi∈Rm
Proof
Let z1,z2∈S,θ≥0. Then there exists x1,x2∈Rn,y11,y12,y21,y22∈Rm such that
Since θx1+(1−θ)x2∈S1,θ(y11+y12),(1−θ)(y21+y22)∈S2,θz1+(1−θ)z2∈S.
Therefore, S is a convex set.
Cartesian product
The Cartesian product of two convex sets
is a convex set.
Proof
Let z1,z2∈S1×S2,θ≥0. Then there exists x1,x2∈S1,y1,y2∈S2 such that
Since θx1+(1−θ)x2∈S1,θy1+(1−θ)y2∈S2, θz1+(1−θ)z2∈S1×S2.
Therefore, S1×S2 is a convex set.
Perspective function
Define the perspective function P:Rn+1→Rn, with domain dom P=Rn×R++, as
where R++ denotes the set of positive numbers. The perspective function scales or normalizes vectors so the last component is one, and then drops the last component.
If C⊂dom P is convex, then its image
is convex.
Proof
Let x,y∈P(C),θ≥0. Then there exists z1,z2∈Rn,t1,t2∈R such that
We want to find μ such that
Since μ≥0 and C is a convex set,
Therefore, P(C) is a convex set.
We can interpret the relation between μ and θ like this. Since two triangulars are similar,
By using this, we can easily derive
We can also show that the line segments are mapped to line segments under the perspective function. The result can be interpreted as a concept of a pin-hole camera.
The inverse image of a convex set under the perspective function is also convex.
If C⊂Rn is convex, then
is convex.
Proof
Let (x1,t1),(x2,t2)∈P−1(C),θ≥0.
We want to show that θ(x1,t1)+(1−θ)(x2,t2)∈P−1(C), i.e.
We want to find μ≥0 such that
Take
Since x1/t1,x2/t2∈C and C is a convex set,
Therefore, P−1(C) is a convex set.
Linear-fractional functions
A linear-fractional function is formed by composing the perspective function with an affine function. Suppose g:Rn→Rm+1 is affine, i.e.
where A∈Rm×n,b∈Rm,c∈Rn, and d∈R. The function f:Rn→Rm given by f=P∘g such that
where dom f={x∣cTx+d>0}
is called a linear-functional (or projective) function. If c=0 and d>0, the domain of f is Rn, and f is an affine function.
Example
Generalized inequalities
Proper cones and generalized inequalities
A cone K⊂Rn is called a proper cone if it satisfies the following
K is convex
K is closed
K is solid (it has nonempty interior)
K is pointed, which means that it contains no line (or equivalently, x∈K,−x∈K⇒x=0)
Examples
nonnegative orthant K=R+n={x∈Rn∣xi≥0,i=1,…,n}
positive semidefinite cone K=S+n
nonnegative polynomials on [0,1]
A proper cone K can be used to define generalized inequality, which is a partial ordering on Rn that has may of the properties of the standard ordering on R defined by
Actually, we have to check whether the above definition satisfy the following conditions
For all a,b,c∈Rn,
Partial orders
Reflexivity : a≤a (every element is related to itself)
Antisymmetry : if a≤b and b≤a then a=b (no two distinct elements precede each other)
Transitivity : if a≤b and b≤c then a≤c
Strict partial orders
Ir-reflexivity : not a<a (no element is related to itself)
Asymmetry : if a<b then not b<a
Transitivity : if a≤b and b≤c then a≤c
Examples
component-wise inequality (K=R+n)
matrix inequality (K=S+n)
Properties
many properties of ⪯K are similar to ≤ on R
⪯K is preserved under addition
Proof
y−x∈K and v−u∈K. Since K is a proper cone,
⪯K is preserved under non-negative scaling
Proof
y−x∈K. Since α>0α>0K is a proper cone,
⪯K is preserved under limits
If xi⪯Kyi for i=1,2,… and xi→x and yi→y as i→∞, then x⪯Ky
Proof
yi−xi∈K,∀i and yi−xi→y−x as i→∞
Since K is closed (it contains all of its limit points),
Therefore, x⪯Ky
Minimum and minimal elements
A point x∈S is the minimum element of S with respect to ⪯K if
If a set has a minimum element, then it is unique.
Proof
Let x and y are minimum element of S. Then,
Therefore,
By anti-symmetry,
Therefore, it is unique.
A point x∈S is the minimal element of S, if
Example
When K=R+2,
x1 is the minimum element of S1
x2 is a minimal element of S2 since the yellow regions are not comparable to x2.
Separating hyperplane
If C and D are disjoint convex sets, then there exists a=0,b such that
the hyperplane {x∣aTx=b} separates C and D
strict separation requires additional assumptions (e.g. C is closed, D is a singleton)
Supporting hyperplane theorem
supporting hyperplane to set C at boundary point x0:
where a=0,∀x∈C,aTx≤aTx0
supporting hyperplane theorem : if C is convex, then there exists a supporting hyperplane at every boundary point of C
Dual cones and generalized inequalities
Let K be a cone. The set
is called the dual cone of K. As the name suggests, K∗ is a cone, and is always convex, even when the original cone K is not.
Proof
Let y1,y2∈K∗,θ≥0.
Therefore, y1Tx≥0 and y2Tx≥0 for all x∈K
We want to show that θy1+(1−θ)y2∈K∗ i.e.
Since y1Tx≥0 and y2Tx≥0 for all x∈K,
So, θy1+(1−θ)y2∈K∗
Therefore, K∗ is a convex set.
Geometrically, y∈K∗ if and only if −y is the normal of a hyperplane that supports K at the origin.
Examples
K=R+n,K∗=R+n
Dual of a norm cone. Let ∥⋅∥ be a norm on Rn. The dual of the associated cone K={(x,t)∈Rn+1∣∥x∥≤t} is the cone defined by the dual norm
where the dual norm is given by ∥u∥∗=sup{uTx∣∥x∥≤1}
Proof
We want to show that
TBC
Note that
Properties
K∗ is closed and convex
K1⊂K2⇒K2∗⊂K1∗
If K has non-empty interior, then K∗ is pointed
If the closure of K is pointed then K∗ has non-empty interior.
K∗∗ is the closure of the convex hull of K (Hence if K is convex and closed, K∗∗=K)
Dual generalized inequalities
Suppose that the convex cone K is proper, it induces a generalized inequality ⪯K. Then its dual cone K∗ is also proper, and therefore, induces a generalized inequality.
Example : positive semi-definite cone
On the set of symmetric n×n matrices Sn, we use the standard inner proudct tr(XY)=∑i,jXijYij. The positive semi-definite cone S+n is self-dual. i.e. for X,Y∈Sn
Proof
Suppose Y∈/S+n. Then there exists y∈Rn with
Since X=qqT is a positive semi-definite matrix, Y∈/(S+n)∗
Now suppose X,Y∈S+n. We can express X as a linear combination of rank-one matrix by using eigenvalue decomposition.
Then,
Therefore, Y∈(S+n)∗
Minimum and minimal elements via dual inequalities
We can use dual generalized inequalities to characterize minimum and minimal elements of a set S⊂Rn with respect to the generalized inequality induced by a proper cone K
Minimum element w.r.t ⪯K
x is a minimum element of S if and only if for all λ≻K∗0, x is the unique minimizer of λTz over z∈S. Geometrically, this means that for any λ≻K∗0, the hyperplane
is a strict supporting hyperplane to S at x. Equivalently, this means that the hyperplane intersects S only at the point x
Minimal element w.r.t ⪯K
If x minimizes λTz over S for some λ≻K∗0, then x is minimal
If x is a minimal element of a convex set S, then there exists a nonzero λ⪰K∗0 such that x minimizes λTz over S.
If S is not a convex, we can’t guarantee that there exists such λ
In this case, x is a minimal point of S with respect to R+2. However, there doesn’t exist λ for which x minimizes λTz over z∈S