Professional Documents
Culture Documents
Slide-1 Convex Set
Slide-1 Convex Set
Applications
Topic 1 - Convex Set
Li Li
Department of Automation
Tsinghua University
Fall, 2009-2022.
-1-
1.0. Outline 纲要
1.-1. Preliminary 预备知识
-2-
1.-1. Preliminary 预备知识
i1 i ,写成向
n
量,以第二范数/欧几里得范数为例, x 2
2
x
2 T
x
量形式为 2 x x。
非负性:对于所有 x ,有 f ( x) 0
n
正定性:当且仅当 x 0 的时候 f ( x) 0
齐次性:对于所有 x n
, t , f (tx) | t | f ( x)
三角不等式:对于所有 x, y n
, f ( x y ) f ( x) f ( y )
-3-
1.-1. Preliminary 预备知识
mn
仿射子空间(affine subspaces):给定矩阵 A 以及向
量b m
,仿射子空间为集合{x n
: Ax b}
仿 射 变 换 ( affine transformation ): 从 n 到 m 的 变 换
y A x c被称为仿射变换,其中 y m ,A mn ,c m 为
常向量
仿射函数(affine function)
:仿射函数为仿射变换的一种特
殊情況,当仿射变换中的 m 1时, AT n , y, c 都为常数;
如果进一步,将此处 y Ax c 中的截距 c 去掉,剩下的部分
y Ax 被称为线性函数(linear function)
-4-
1.1. Definition and Examples 定义和例子
为什么我们从凸集开始
布尔巴基学派认为“数学结构勾画了数学体系内部的关系,
数学结构简称为结构,结构由若干集合,定义在集合上或集
合间的一些关系,以及一组作为条件的公理组成”
--《数学辞海》编辑委员会《数学辞海(第一卷)》
希望大家在学习的过程中,能逐渐熟悉相对抽象的性质和
定理,同时又掌握足够多的相对具象的实例
-5-
1.1. Definition and Examples 定义和例子
点Point,在我们课中,如无特殊说明,指欧式空间中的点
-7-
1.1. Definition and Examples 定义和例子
清华大学《凸优化》课程 2022年 作业题
-8-
1.1. Definition and Examples 定义和例子
凸集 Convex Set,Set C is a convex set, if for any x, y C ,
and 0,1 , we have x 1 y C
典型的凸集包括:
点 Point,线 Line,射线 Ray,面 Hyperplane,半空间 Half
Space
锥 Cone
球 Ball,多面体 Polyhedral P1 {x | Ax b}
-9-
1.1. Definition and Examples 定义和例子
球 Ball、超平面 Hyperplane、半平面 Half Space 都是凸集
作为超平面、半平面的交集的可行域是多面体 Polyhedron,
多面体是凸集
- 10 -
1.1. Definition and Examples 定义和例子
球 Ball,开球 OpenBall,闭球 Closed Ball
- 11 -
1.1. Definition and Examples 定义和例子
开球和闭球都是凸集
- 12 -
1.1. Definition and Examples 定义和例子
超平面 Hyperplane
- 13 -
1.1. Definition and Examples 定义和例子
超平面比原空间的维度要低一维
半平面 Halfspace
- 14 -
1.1. Definition and Examples 定义和例子
半平面 Half Space 是凸集
- 15 -
1.1. Definition and Examples 定义和例子
边界点 Boundary Point
- 16 -
1.1. Definition and Examples 定义和例子
顶点/极点Extreme Point,如果凸集内的一点不在凸集内任
何不同的两点的连线上,则称该点为该凸集的顶点
- 17 -
1.1. Definition and Examples 定义和例子
多面体Polyhedron
凸集的交集也是凸集
- 18 -
1.1. Definition and Examples 定义和例子
- 19 -
1.1. Definition and Examples 定义和例子
多面体Polyhedron是凸集可以用下面的方法来证明:
大前提:线性规划问题的约束条件为若干个线性等式或者
不等式,而这些集合都是凸集
小前提:凸集的交集也是凸集
结论:作为这些凸集交集的线性规划问题定义域也是凸集
(我们约定空集为凸集)
- 20 -
1.1. Definition and Examples 定义和例子
锥 Conic (nonnegative) combination of x1 and x 2 : any point
of the form x 1 x1 2 x2 , with 1 , 2 0 .
- 21 -
1.1. Definition and Examples 定义和例子
锥 Cone 的性质
A set n
is a cone if
x x , 0
- 23 -
1.1. Definition and Examples 定义和例子
尖锥 Pointed Cone 和多面体锥 Polyhedral Cone
- 24 -
- 25 -
1.1. Definition and Examples 定义和例子
- 27 -
1.1. Definition and Examples 定义和例子
我们可以证明这两个定义的等价性
- 28 -
- 29 -
- 30 -
- 31 -
1.1. Definition and Examples 定义和例子
- 32 -
1.1. Definition and Examples 定义和例子
单纯形Simplex:We call the convex hull of any set of n 1
points in R n which do not lie on a hyperplane a simplex.
- 33 -
1.1. Definition and Examples 定义和例子
半正定矩阵Positive Semidefinite Matrix
n
The set of real symmetric n n matrices is denoted S . A
matrix A S is called positive semidefinite if x Ax 0 for
n T
n
S
denoted and the set of positive definite matrices is denoted
n n
S S
by . is a closed, convex, pointed, and solid cone.
2) for all x R , x Ax 0
n T
- 35 -
1.1. Definition and Examples 定义和例子
Proof: 1) and 2) are equivalent by definition. 3) and 4) are
obvious according to Lemma 1.1 below.
Corollary 5.1: Every psd (pd) matrix A has a unique psd (pd)
square root A1 / 2 such that A1/ 2 A1/ 2 A and every pd matrix is
nonsingular.
- 36 -
1.1. Definition and Examples 定义和例子
Proposition 1.2: The following statements are equivalent:
n
1) matrix A S is positive
2) for all x R , x Ax 0
n T
- 37 -
1.1. Definition and Examples 定义和例子
Lemma 1.1: Any a real Hermitian matrix A can be
diagonalized by an orthogonal matrix.
Proof:
1) all the eigenvalues and eigenvectors are real, notice that
( A) ( AT )
2) the eigenvectors of different eigenvalues are orthogonal.
Suppose 1 1 X AX 1, X
2 2 AX 2, 1 2, X 1 X 2 0 , because
T
- 38 -
1.1. Definition and Examples 定义和例子
4) based on 2) and 3), we can see that A has n linear
independent eigenvectors. Via Gram-Schmidt orthogonalization,
we can make these eigenvectors orthogonal. So it is proved.
Another Proof:
According to the Schur triangularization theorem there is a
unitary matrix U such that U AU R , where R is upper
T
- 39 -
1.1. Definition and Examples 定义和例子
半正定锥 Positive Semidefinite Cone:The set of all PSD
matrices A Rnn forms a convex cone.
x y
A
For example, let y z , we have
- 40 -
1.1. Definition and Examples 定义和例子
Is the following set convex?
x1 0 , x1 x2 1
- 41 -
1.1. Definition and Examples 定义和例子
Yes
x1 1
1 0
x2
- 42 -
1.1. Definition and Examples 定义和例子
Is the following set convex?
x2 x12
- 43 -
1.1. Definition and Examples 定义和例子
Yes
1 x1
x 0
1 x2
- 44 -
1.1. Definition and Examples 定义和例子
Is the following set convex?
x12 x22 1
- 45 -
1.1. Definition and Examples 定义和例子
Yes
1 x1 x2
x 0
2 1 x1
- 46 -
1.1. Definition and Examples 定义和例子
Is the following set convex?
x2 x12 or x1 0 , x2 0
- 47 -
1.1. Definition and Examples 定义和例子
Yes, but cannot be written into a SDP (LMI) form, because
not basic semialgebraic (a semialgebraic set is a subset S of
real n -dimensional space defined by a finite sequence of
polynomial equations and inequalities; or any finite union of
such sets.)
- 48 -
1.1. Definition and Examples 定义和例子
Is the following set convex?
- 49 -
1.1. Definition and Examples 定义和例子
No because not connected
- 50 -
1.1. Definition and Examples 定义和例子
Is the following set convex?
- 51 -
1.1. Definition and Examples 定义和例子
Yes
1 x1 0
x 1 x2 0
1
0 x2 1 2 x1
- 52 -
1.1. Definition and Examples 定义和例子
Is the following set convex?
x14 x24 1
- 53 -
1.1. Definition and Examples 定义和例子
Yes, but cannot be written into a SDP (LMI) form.
- 54 -
1.1. Definition and Examples 定义和例子
半正定矩阵和椭球/椭球面 Ellipse:Any a PSD matrix A is
corresponding to a special ellipse x Ax I .
T
- 55 -
1.1. Definition and Examples 定义和例子
清华大学《凸优化》课程 2021年 期末考试题
n n
如果一个定义在 R 上的集合 C 和任意一条 R 中的直线的
交集是凸集,请问该集合是否是凸集?并给出理由。
- 56 -
1.2. Generalized Inequalities 广义不等式
布尔巴基学派将数学结构分为三个元结构,分别是序结构、
代数结构和拓扑结构。其中序结构即由集合及在其上规定的
序关系/偏序关系Partial Ordering组成的数学结构。
- 57 -
1.2. Generalized Inequalities 广义不等式
特别对于一些凸集,我们也可以定义相应的序结构。比如
对于凸锥,有了射线而不是直线,则有了方向!
Every proper cone K in R n induces a partial ordering K
defining 广义不等式generalized inequalities on R n
a K b a b K
- 59 -
1.3. Operations Preserve Convexity 保凸运算
除了上述常见实例,如何分析手头的集合是否凸集?
- 60 -
1.3. Operations Preserve Convexity 保凸运算
- 61 -
1.3. Operations Preserve Convexity 保凸运算
- 62 -
1.3. Operations Preserve Convexity 保凸运算
设 A mn , 证 明 如 果 S m 是 一 个 凸 集 , 那 么
A1 ( S ) {x n : Ax S}也是一个凸集。
A1 ( S ) 一般称为原象集(pre-image)
。
- 63 -
1.3. Operations Preserve Convexity 保凸运算
,那么所有满足不等式 i 1 xi Ai
k
给定 A1 , , Ak , B n
B的
点构成的集合为凸集。
这个问题要想解释清楚,使用常规的凸集证明方式即可,
但是有一种更加有趣的证明方法。
, f ( x) B i 1 xi Ai 那么这个时候我们会发
k
设f : k n
现
k
f ( ) {x : f ( x) } x : B xi Ai 0
1 n n
i 1
而这个就是我们关注的那个集合。这个集合当然是凸集,
因为它是一个凸集在线性映射下的原象集。
- 64 -
1.3. Operations Preserve Convexity 保凸运算
- 65 -
1.4. Separating Hyperplane 凸集分离
分离超平面
- 66 -
1.4. Separating Hyperplane 凸集分离
支撑超平面
- 67 -
1.4. Separating Hyperplane 凸集分离
- 68 -
1.4. Separating Hyperplane 凸集分离
- 69 -
1.4. Separating Hyperplane 凸集分离
- 70 -
1.4. Separating Hyperplane 凸集分离
- 71 -
1.4. Separating Hyperplane 凸集分离
- 72 -
1.4. Separating Hyperplane 凸集分离
- 73 -
1.4. Separating Hyperplane 凸集分离
- 74 -
1.4. Separating Hyperplane 凸集分离
超平面分离定理:Suppose C and D are two convex sets in
R n that do not intersect, i.e., C D . Then there exist a
a Rn , a 0 and b R such that aT x b for all x C and
aT x b for all x D . In other words, the affine function aT x b
is nonpositive on C and nonnegative on D . The hyperplane
x | aT x b is called a separating hyperplane for the sets C and
D , or is said to separate the sets C and D .
,
C x R2 | x1 0
D x R2 | x1 x2 1
- 76 -
1.5. Some Quiz Problems 测试问题
For any a convex 2n -sided polyhedron in a R 2 plane.
Suppose we color any n vertices into red, the rest n vertices
into blue. Please prove or disprove there must be a point on the
plane so that the sum of its distance from all the red points is
equal to the sum of its distance from all the blue points.
- 77 -
1.5. Some Quiz Problems 测试问题
只需要证明离给定点最近的那条边的确满足投影点在边上
的要求就可以了。假设上图中的红色虚线是给定点到所有边
的垂线段中最短的一个,但垂足却在边的外面。我们立即发
现,存在某个边对应的蓝色垂足(请问为什么?),由于灰
色直角三角形中斜边大于直角边,蓝色垂线段显然要比红色
线段更短(注意:蓝线的垂线不一定要求落在边上),这就
与红色线段是所有垂线段中最短相矛盾。
这个命题对更高维的情形也都是成立的:对于给定凸多面
体和它内部的一点,总能找到其中一个面使得,给定点在这
个面上的投影恰好就落在这个面上。
- 79 -
1.5. Some Quiz Problems 测试问题
A Gömböc is a convex three-dimensional homogeneous body
which, when resting on a flat surface, has just one stable and
one unstable point of equilibrium. Its existence was conjectured
by Russian mathematician Vladimir Arnold in 1995 and proven
in 2006 by Hungarian scientists Gábor Domokos and Péter
Várkonyi. The Gömböc shape is not unique; it has countless
varieties, most of which are very close to a sphere and all have
very strict shape tolerance (about 0.1 mm per 10 cm). The most
famous solution has a sharpened top and is shown on the right.
Its shape helped to explain the body structure of some turtles in
relation to their ability to return to equilibrium position after
being placed upside down.
- 80 -
1.5. Some Quiz Problems 测试问题
- 81 -
1.5. Some Quiz Problems 测试问题
Suppose you are given a collection of squares and equilateral
triangles (of unit side). You are asked to form convex polygons
by sticking the squares and triangles together with their sides
aligned. These polygons must also have unit sides (so that for
example sticking two squares together to form a domino does
not count). How many distinct polygons can be formed in this
way (including polygons formed by a single square or triangle)?
- 82 -
- 83 -
1.5. Some Quiz Problems 测试问题
我们的学习要求:
了解基本概念
形成足够直觉
完成简单题目(需要综合分析,代数和拓扑的基础知识)
我们是有选择性的介绍部分定理的详细证明,目的有两个:
其一,熟悉相关概念的性质;其二,学习相关的证明思路,
提高数学能力。我们的考试一般不涉及证明。
- 84 -
1.7. References
[1] S. Boyd, L. Vandenberghe, Convex Optimization, Cambridge
University Press, 2004. http://www.stanford.edu/~boyd/cvxbook/
http://www.ee.ucla.edu/~vandenbe/cvxbook
[2] D. Bertsimas, J. N. Tsitsiklis, Introduction to Linear Optimization,
Athena Scientific, 1997.
[3] Gömböc http://en.wikipedia.org/wiki/G%C3%B6mb%C3%B6c
[4] http://domino.research.ibm.com/Comm/wwwr_ponder.nsf/Challenges/
April2008.html
- 85 -