• No results found

Quantum and Non-deterministic computers facing NP-completeness

N/A
N/A
Protected

Academic year: 2021

Share "Quantum and Non-deterministic computers facing NP-completeness"

Copied!
34
0
0

Loading.... (view fulltext now)

Full text

(1)

facing NP-completeness

Thibaut Barthelemy

University of Vienna Dept. of Business Administration

Austria

Vienna January 29th, 2013

Some pictures come from Wikipedia

(2)

Introduction

Slide 2 / 21 Barthelemy

(3)

Introduction

Contents

Introduction

Complexity of some problems for some computers

Do NP-hard problems become polynomial on a quantum computer ?

Basics of Complexity Theory

Deterministic and Non-deterministic Turing machines Definition of class NP

Definition of NP-completeness

SAT complexity for 3 kinds of computers

How a Deterministic computer solves SAT How a Quantum computer solves SAT How a DNA computer solves SAT

(4)

Introduction

Overview

Complexity of some problems for some computers

Class Deterministic Quantum Non deterministic

Search P O(n)1 O(√

n)2 O(log n)3 Factorization N P O(n1.73n)4 O(n3)5 O(n)3

SAT N P-C O(2n)1 O(√

2n)2 O(n)6

1. Enumeration 2. Grover’s algorithm

3. Recursively search both half-parts in parallel 4. General Number Field Sieve algorithm 5. Shor’s algorithm

6. Recursive enumeration of variables for both values in parallel

Slide 4 / 21 Barthelemy

(5)

Introduction

Quantum power

Gain for factorization

Factorizing a huge integer is

exponential on classical computers, cubic on quantum computers.

But this problem is specific :

it can be solved by means of quantum Fourier transform, which is very fast ;

it has never been proven to be NP-hard.

Gain for NP-hard problems ?

No polynomial algorithm has been found for any NP-complete problem : SAT, although simple, remains intractable : O(√

2n) = O(2n/2) All NP-complete problems might remain intractable

(6)

Basics of

Complexity Theory

Slide 6 / 21 Barthelemy

(7)

Complexity Theory

Definitions

The Turing Machine

(8)

Complexity Theory

Definitions

Church-Turing thesis (20th century)

Roughly,

a Turing Machine can compute every problem for which we know a solving way ;

equivalently, every algorithm can be translated into a TM transition table.

Slide 8 / 21 Barthelemy

(9)

Complexity Theory

Definitions

The Deterministic Turing Machine

Deterministic means both

it cannot guess the right choice, it is in one state at once.

So the solution tree of Hamiltonian Cycle is explored branch by branch :

(10)

Complexity Theory

Definitions

The Non-deterministic Turing Machine

Non-deterministic means either it can be in several states at once, or it can guess the right choice.

So, at a given level, either

for every node, in parallel, it deploys its branches ; from the correct node, it chooses the next correct node.

Slide 10 / 21 Barthelemy

(11)

Complexity Theory

Theorem

Cook-Levin theorem

This theorem says, equivalently

every problem in NP can be reduced to SAT in polynomial time, SAT is NP-complete.

Decision problems. Whenever the name evokes an optimization problem, its decision variant is considered.

(12)

Complexity Theory

Theorem

Cook-Levin theorem : proof sketch

1. SAT is in NP (means Polynomial on a Non-deterministic machine).

2. Given the transition table of any problem of size n in NP, the corresponding behavior of the Turing Machine can be translated into a SAT problem whose the size is polynomial with respect to n, 3. so SAT can solve any problem of NP.

Slide 11 / 21 Barthelemy

(13)

Complexity Theory

NP-completeness

Other NP-complete problems

Many problems are also NP-complete, because

they can be used to solve SAT (SAT is reducible to them in polynomial time).

(14)

Complexity Theory

NP-completeness

Other NP-complete problems

Many problems are also NP-complete, because

they can be used to solve SAT (SAT is reducible to them in polynomial time).

Slide 12 / 21 Barthelemy

(15)

SAT complexity

for 3 kinds of computers

(16)

SAT complexity

On deterministic computers

Exploration

The worst case of the best algorithms is equivalent to trying both values for all variables

Slide 14 / 21 Barthelemy

(17)

SAT complexity

On quantum computers

Implementing the instance

First,

build the quantum circuit of the instance,

it takes (x1, . . . , xn) and gives f (x) = 1 iff all clauses are satisfied

(18)

SAT complexity

On quantum computers

Making the solution emerge

Use the Grover’s search algorithm with f (xα) as certifier.

αx← |0i − |1i

√2 ∀x ∈ {0 . . . 2n− 1}

xα ← 1

√2n

2n−1

X

x=0

|xiαx

for k = 1 to√ 2n do xα ←

2n−1

X

x=0

(−1)f (xα)|xiαx

xα ←

2n−1

X

x=0

|xi(2α − αx)

Slide 16 / 21 Barthelemy

(19)

SAT complexity

On quantum computers

Making the solution emerge

Repeat the Grover’s iteration√

2n times :

(20)

SAT complexity

On quantum computers

Making the solution emerge

Repeat the Grover’s iteration√

2n times :

Slide 17 / 21 Barthelemy

(21)

SAT complexity

On quantum computers

Making the solution emerge

Repeat the Grover’s iteration√

2n times :

(22)

SAT complexity

On quantum computers

Making the solution emerge

Repeat the Grover’s iteration√

2n times :

Slide 17 / 21 Barthelemy

(23)

SAT complexity

On DNA computers

Biochemistry basics : DNA, Genes

(24)

SAT complexity

On DNA computers

Biochemistry basics : Separation, Hybridization

Slide 18 / 21 Barthelemy

(25)

SAT complexity

On DNA computers

Biochemistry basics : Electrophoresis

A voltage makes molecules move The lightest ones go far

(26)

SAT complexity

On DNA computers

Encoding

Any 15-bases sequence is a variable

The sequence is unique for each variable and value

Braich, Chelyapov, Johnson, Rothemund, Adleman

Solution of a 20-Variable 3-SAT Problem on a DNA Computer Science Vol. 296 no. 5567 (2002)

Slide 19 / 21 Barthelemy

(27)

SAT complexity

On DNA computers

Generation of all solutions

O(2n) solutions are generated in n steps :

(28)

SAT complexity

On DNA computers

Extraction of feasible solutions

Solutions are poured on the left layer

Base-complements of Clause 1 variables are stuck on the right

Slide 20 / 21 Barthelemy

(29)

SAT complexity

On DNA computers

Extraction of feasible solutions

Solutions move to the right

The right layer catches DNA matching at least one variable

(30)

SAT complexity

On DNA computers

Extraction of feasible solutions

The left layer is replaced by the right layer

Base-complements of Clause 2 variables are stuck on the right

Slide 20 / 21 Barthelemy

(31)

SAT complexity

On DNA computers

Extraction of feasible solutions

Solutions move to the right

The right layer catches DNA matching at least one variable

(32)

SAT complexity

On DNA computers

Extraction of feasible solutions

The left layer is replaced by the right layer

Base-complements of Clause 3 variables are stuck on the right

Slide 20 / 21 Barthelemy

(33)

SAT complexity

On DNA computers

Extraction of feasible solutions

Solutions move to the right

The right layer catches DNA matching at least one variable

(34)

SAT complexity

On DNA computers

Complexity of that algorithm

There are m iterations during search, but

after each iteration, one may need to duplicate the remaining strands ;

a chemical procedure called PCR makes O(2n) duplicates in O(n).

Thus the overall complexity is O(mn).

Drawbacks

Non-deterministic computing swaps time and space complexities : Very fast : O(mn)

Matter-consuming : O(2n)

The size of the Universe becomes our limit, instead of its lifetime

Slide 21 / 21 Barthelemy

References

Related documents