证明与计算(1): Decision Problem, Formal Language L, P and NP

0x01 从判定问题到形式语言

这篇讲知识证明的wiki([1]):

https://en.wikipedia.org/wiki/Proof_of_knowledge

里面有一句话:

Let x be a language element of language L in NP

这篇讲NPC的文章([2])

http://www.cs.princeton.edu/courses/archive/spr11/cos423/Lectures/NP-completeness.pdf

里面提到Decision Problem([4)，Decision Problem就是判定yes/no的问题，一个Decision Problem Q可以等价于一个Formal language L：

L = {x ∈ {0,1}* : Q(x) = 1}

资料[4]里举例了Graphs, Deterministic Finite Automaton, DFA问题，Context Free Grammar, CFG问题，Turing Machine问题，素数分解，Boolean satisfiability problem等等都可以通过编码成字符串，进而转成形式语言。

例如，下面是一个上下文无关语言的产生式：

• S->XSY|e
• X->aX|a
• Y->bY|e

可以编码成：

• a list of its variables
• a list of its terminal symbols
• a list of its rules as (LHS, RHS) pairs
• its start variable

那么，转成字符串就是：

• '<G>=(S,X,Y)(a,b)((S,XSY),(S,e),(X,aX),(X,a),(Y,bY),(Y,e))S'

实际上，字符串二进制化后就变成了只用两个字符({0,1})表示的字符串:

const decesionProblem = '<G>=(S,X,Y)(a,b)((S,XSY),(S,e),(X,aX),(X,a),(Y,bY),(Y,e))S'.split('')
const reducer =
  (accumulator, currentValue) =>
    accumulator + currentValue.charCodeAt(0).toString(2);
const binaryString = decesionProblem.reduce(reducer,'');
console.log(binaryString);

输出：

"11110010001111111101111011010001010011101100101100010110010110011010011010001100001101100110001010100110100010100010100111011001011000101001110110011010011011001010001010011101100110010110100110110010100010110001011001100001101100010100110110010100010110001011001100001101001101100101000101100110110011000101011001101001101100101000101100110110011001011010011010011010011"

可以理解为Q是判定算法，该判定算法能确定一个具体的Decision Problem，其结果是yes还是no。

那么，L是一个集合，其中每个x都是被Q判定为yes的Decison Problem的实例。

那么，P问题等价于如下的语言集合：

P = {L \subseteq {0, 1}* : exist an algorithm A that decides L in p-time}

我们可以理解为，P是由所有能在多项式时间内确定yes/no的Q所导出的语言L组成。是不是很绕？分解一下：

• Q是一个判定算法。
• Q是一个能在多项式时间内判定yes/no的判定算法。
• L是被Q判定结果为yes的所有输入x的集合。
• L是被Q判定结果为yes的所有输入x的集合所表示的形式语言。

0x02 P问题

在[3]里面描述：

P is the set of languages whose memberships are decidable by a Turing Machine that makes a polynomial number of steps.

这里增加了Turing Machine的限定，也就是Q是在Turing Machine上执行的。参考链接[5],[6]里是对确定性图灵机和非确定性图灵机的描述。

例如，下面的k路径问题，属于P ([2])：

PATH = {< G, u, v, k > : G = (V, E) is an undirected

graph, u,v ∈ V, k ≥ 0 is an integer, and exist a path

from u to v in G with ≤ k edges}

因此，P是一类判定问题的集合，同时等价于一类形式语言的集合，这些语言在确定性图灵机上存在多项式(Polynomial)时间复杂度算法。

0x03 证明凭据(certificate)，证明(certifier)

算法A验证了语言L，iff([2])：

L = {x ∈ {0, 1}* : exist y ∈ {0, 1}* s.t. A(x, y) = 1}

通过例子理解Certificate，例如下面的Independent Set属于NP([3])：

Given a graph G, is there set S of size ≥ k such that no two nodes in S are connected by an edge?

• Finding the set S is hard
• But if I give you a set S∗, checking whether S∗ is the answer is easy
• S∗ acts as a certificate that ⟨G,k⟩ is a yes instance of Independent Set

Certificate是否有效(efficient)([3])：

An algorithm B is an efficient certifier for problem X if:

• B is a polynomial time algorithm that takes two input strings I (instance of X) and C (a certificate)
• B outputs either yes or no.
• There is a polynomial p(n) such that for every string I: I ∈ X if and only if there exists string C of length ≤ p(|I|) such that B(I,C) = yes.

Certification的含义是“帮手”([3])：

B is an algorithm that can decide whether an instance I is a yes instance if it is given some “help” in the form of a polynomially long certificate.

注意，C是Certificate，而算法B是certifier

Let’s say you had an efficient certifier B for the Independent Set problem.

怎样找到Certificate呢？

Try every string C of length ≤ p(|I|) and ask is B(I,C) = yes?

0x04 NP问题

那么，NP问题等价于如下的语言集合([2])：

NP = {L \subseteq {0, 1}* : exist a certificate y, |y| = O(|x|^k), and an algorithm A s.t. A(x, y) = 1}

在[3]里面描述NP：

NP is the set of languages for which there exists an efficient certifier.

区别于P：

P is the set of languages for which there exists an efficient certifier that ignores the certificate.

就是说P和NP问题都能在多项式时间内判定，但是NP问题的判定需要certificate的帮助:

A problem is in P if we can decided them in polynomial time. It is in NP if we can decide them in polynomial time, if we are given the right certificate.

我们可以理解为，如果：

• 存在长度是字符串x的多项式倍的字符串y
• 存在验证算法A
• 使得A(x,y)=1

那么：

• 能通过A(x,y)=1的所有x构成的语言是L
• 所有L的集合是NP语言集合

例如，下面的子集求和问题，属于NP：

SUBSET-SUM: Given finite set S of integers, is there a subset whose sum is exactly t?

因此，P问题是一类判定问题的集合，同时等价于一类形式语言的集合，这类语言在确定性图灵机上存在依赖于Certificate的多项式(Polynomial)时间复杂度的验证算法(Certifier)，在非确定性图灵机上(例如量子计算机)存在多项式时间复杂度的求解算法。

0x05 P vs NP vs NPC

中间讨论了P、NP、NPC问题，简单说：

• P \subseteq NP \subseteq NP-hard
• Co-NP = {L': L ∈ NP}
• NPC=NP中最难的集合，并且他们等价

证明P \subseteq NP是很容易的：

Proof. Suppose X ∈ P. Then there is a polynomial-time algorithm A for X.

To show that X ∈ NP, we need to design an efficient certifier B(I,C).

JusttakeB(I,C)=A(I).

0x06 夹逼法(Reduction and NPC Reduction)

下面是对L语言进行规约，两个步骤：

Reduce language L1 to L2 via function f:

1. Convert input x of L1 to instance f(x) of L2
2. Apply decision algorithm for L2 to f(x)

则L2的判定时间>=L1的判定时间，即：L1≤L2

引入p≤的概念：

L1is p-time reducible to L2, or L1 p≤ L2, if exist a ptime

computable function f : {0, 1}* -> {0, 1}* s.t. for all x ∈ {0, 1}*, x ∈ L1

iff f(x) ∈ L2

从而，要判定一个形式语言是否是P的，使用夹逼法：

If L1 p≤ L2 and L2 ∈ P, then L1 ∈ P

进一步，要判定一个形式语言是否是NPC的，使用夹逼法：

A language L ∈ {0, 1}* is NP-complete if:

1. L ∈ NP, and
2. L’ p≤ L for every L’ ∈ NP, i.e. L is NP-hard

利用其他已知的NPC问题，就可以夹逼：

1. If L is language s.t. L’ p≤L where L’ ∈ NPC, then L is NP-hard.
2. If L ∈ NP, then L ∈ NPC.

这样就得到证明NPC的步骤：

This gives us a recipe for proving any L ∈ NPC:

1. Prove L ∈ NP
2. Select L’ ∈ NPC
3. Describe algorithm to compute f mapping every input x of L’ to input f(x) of L
4. Prove f satisfies x ∈ L’ iff f(x) ∈ L, for all x ∈ {0, 1}*
5. Prove computing f takes p-time

最后，如果P=NP，那么....:

“If P = NP, then the world would be a profoundly different place than we usually assume it to be. There would be no special value in "creative leaps," no fundamental gap between solving a problem and recognizing the solution once it's found. Everyone who could appreciate a symphony would be Mozart; everyone who could follow a step-by-step argument would be Gauss...”

— Scott Aaronson, MIT

但是证明或者证伪P=NP是很难的。

0x07 wait to be continue...

我们最开始是为了理解在Proof of knowledge ([1]) 里面这句话的作用：

Let x be a language element of language L in NP

我们需要一步一步拆解。

0x08 参考

[1] wiki:Proof_of_knowledge

[2] cs.princeton.edu:NP-completeness.pdf

[3] cs.cmu.edu:np.pdf

[4] ccs.neu.edu:Decision-Problems

[5] wiki:Turing_machine

[6] wiki:Non-deterministic_Turing_machine