Notes on Probability Essentials - 1 - Axioms of Probability

原创 2015年07月07日 10:48:58

原文发布在作业部落:https://www.zybuluo.com/Emptyset/note/123426

“他说 / 你任何为人称道的美丽 / 不及他第一次遇见你。”
——《南山南》

本文用尽可能少的内容提供了建立概率测度(probability measure)所需要的公理和定义。

Abstract:这里想谈一下直觉性的理解,为什么概率测度要与σ-algebra扯上关系呢?就概率事件而言,空集和全集肯定都得在A里头,这刚好是下文的性质1;任何一个事件必然有它的反面存在,这就是性质2——任何一个事件的补集也必须在A里头;性质4也是满足了概率的研究条件——因为概率测度(probability measure)的建立需要countable additivity, 那么显然这性质4(countable union and intersection)就是需要有的。所以σ-algebra就像是为概率测度量身打造的一种代数结构。
这一节介绍了σ-algebra,定义在R上的Borel σ-algebra, probability measure,以证明完概率测度P具有连续性结束。

Let Ω be an abstract space. Let 2Ω denote all subsets of Ω, including the empty set denoted by . With A being a subset of 2Ω, we consider the following properties:

  1. A and ΩA;
  2. If AA then AcA, where Ac denotes the complement of A;
  3. A is closed under finite unions and finite intersections: that is, if A1,...,An are all in A, then ni=1Ai and ni=1Ai are in A as well;
  4. A is closed under countable unions and intersections: that is, if A1,A2,A3,... is a countable sequence of events in A, then ni=1Ai and ni=1Ai are both also in A.

Definition 1 A is an algebra if it satisfies (1), (2) and (3) above. It is a σ-algebra (or a σ-field) if it satisfies (1),(2), and (4) above.
Note: (1)+(4) implies (3), hence any σ-algebra is an algebra (Exercise: Suppose that Ω is an infinite set, countable or not, and let A be the family of all subsets which are either finite or have a finite complement. Show that A is an algebra, but not a σ-algebra).

Definition 2 If C2Ω, the σ-algebra generated by C, and written σ(C), is the smallest σ-algebra containing C.

Example:
1. A={,Ω} (the trivial σalgebra)
2. AΩ, then σ(A)={,A,Ac,Ω}
3. If Ω=R, the Borel σalgebra is the σalgebra generated by the open sets.

Theorem 1 The Borel σ-algebra of R is generated by intervals of the form (,a], where aQ.

Proof: Let C denote all open intervals. Since every open set in R is the countable union of open intervals, we have σ(C)= the Borel σ-algebra of R.
Let D denote all intervals of the form (,a], where aQ. Let (a,b)C, and let (an)n1 be a sequence of rationals decreasing to a and (bn)n1 be a sequence of rationals increasing strictly to b. Then

(a,b)=n=1(an,bn]=n=1((,bn](,an]c)

Therefore Cσ(D), where σ(C)σ(D). However since each element of D is a closed set, it is also a Borel set, and therefore σ(D) is contained in the Borel sets B. Thus we have
B=σ(C)σ(D)B

and hence σ(D)=B.

Definition 3 A probability measure defined on a σ-algebra A of Ω is a function P:A[0,1] that satisfies:
1. P(Ω)=1
2. For every countable sequence (An)n1 of elements of A, pairwise disjoint(that is, AnAm= whenever nm), one has

P(n=1An)=n=1P(An)

Note: Axiom (2) is called countable additivity, the number P(A) is called the probability of the event A.

Theorem 2 If P is a probability measure on (Ω,A), then:
1. P()=0
2. P is finite additive.

Note: 第二点finite additivity可以直接Definition 3(2)的countable additivity获得。需要注意finite additivity并不是countable additivity的充分条件。尽管finite additivity从直觉上很容易理解,但是它能做的事情实在是太少了(比如它无法处理极限情况),下面这个定理将揭示countable additivity的作用。

Theorem 3 Let A be a σ-algebra. Suppose that P:A[0,1] satisfies (1) and is additive. Then the following are equivalent:
1. Axiom(2) of Definition 3 (Countable Additivity)
2. If AnA and An, then P(An)0.
3. If AnA and AnA, then P(An)P(A).
4. If AnA and AnΩ, then P(An)1.
5. If AnA and AnA, then P(An)P(A).

Note:这里的Notation AnA是指AnAn+1 and n=1An=A; AnA是指An+1An and n=1An=A

Theorem 4 Let P be a probability measure, and let An be a sequence of events in A which converges to A. Then AA and limnP(An)=P(A).

Proof: Let us define

lim supnAn=n=1mnAmlim infnAn=n=1mnAm

Since A is a σ-algebra, we have lim supnAnA and lim infnAnA.
By hypothesis An converges to A, which means limn1An=1A, all ω. This is equivalent to saying that A=lim supnAn=lim infnAn. Therefore AA.
Now let Bn=mnAm and Cn=mnAm. Then limnP(Bn)=limnP(Cn)=P(A), by Theorem 3. However BnAnCn, therefore P(Bn)P(An)P(Cn), so limnP(An)=P(A) as well.

Note: 这里通过引入lim suplim inf,将An转化成了一个递增与一个递减的序列。而后就可以直接利用Theorem 3证明结论了。

版权声明:本文为博主原创文章,如需转载请注明来源。

相关文章推荐

概率论札记 - 2 - 用贝叶斯定理来讨论“医疗诊断的可靠性到底有多少”

只有愚蠢的人才会相信眼睛看到的。 ——安·兰德 故事要从一道贝叶斯定理的简单习题讲起。大意是艾滋病患病率为万分之一,误诊率为5%,患有艾滋病者被诊断出来的概率为99%,请问在这样的设定下如果你被...

Study notes for Continuous Probability Distributions

This article describes the basics of probability and the most commonly used continuous probability d...

Probability Essentials

  • 2009-09-09 07:52
  • 13.99MB
  • 下载

Principle of Computing (Python)学习笔记(3) probability +Objects and reference + tic_tac_toe

1 Basic Probability    https://class.coursera.org/principlescomputing-001/wiki/view?page=probabilit...

A Tutorial on Probability Theory

  • 2016-02-26 08:44
  • 259KB
  • 下载

Collision probability of p-stable LSH

LSH是用局部敏感的方法解决近似最近邻搜索的问题。在原始的LSH方法中,通过将原始空间嵌入到Hamming空间中,将d维空间转换成d'=Cd维的Hamming空间(C是指原始空间中点的坐标的最大值,具...

probability applied on engineering

  • 2015-07-16 14:26
  • 13.54MB
  • 下载

Classification Probability Models and Conditional Random Fields(1)--Naive Bayes

目前正在学习自然语言处理相关的概率模型,在一篇名为《Classification Probability Models and Conditional Random Fields》论文中讲述了常用的几...
内容举报
返回顶部
收藏助手
不良信息举报
您举报文章:深度学习:神经网络中的前向传播和反向传播算法推导
举报原因:
原因补充:

(最多只允许输入30个字)