CS 3261 Computability Course Summary

Post on 16-Jan-2016

43 views 2 download

Tags:

description

CS 3261 Computability Course Summary. Zeph Grunschlag. Announcements. Last hw due now Look out for a final exam practice problems coming out over the weekend I will hold final review session on Tuesday 12/11, 3-5 pm, 833 Mudd. Pick-up final hw’s. - PowerPoint PPT Presentation

transcript

1

CS 3261 Computability Course Summary

Zeph Grunschlag

2

Announcements

Last hw due nowLook out for a final exam practice problems coming out over the weekend I will hold final review session on Tuesday 12/11, 3-5 pm, 833 Mudd. Pick-up final hw’s. I will hold daily OH’s next week and Monday 12/17, 12:00-1:30 except Thursday, 12/13 Final exam: Tuesday 12/18, 9-12, 833 Mudd

3

Computability ConceptsAIM: Reduce Computer Science to

its bare theoretical essentials.

APPROACH: Algorithmic Problems

Formal Languages

4

Formal LanguagesFundamental insight of Theoretical CS:By understanding how formal languages can

be computed, will understand how any algorithmic problem can be solved.

Algorithmic input/output problems involve creating procedures for procuring outputs from given inputs. Can be turned into a formal languages by re-writing as yes/no questions.

EG: “Find the shortest path…” becomes“Is there a path shorter than…”

5

Computability ConceptsAIM: Reduce Computer Science to its bare

theoretical essentials.Algorithmic Problems Formal LanguagesComputers Graph based machine models

Questions to investigate: 1) What sorts of problems can be solved by each

computer model? 2) What languages does each model accept?3) What are the practical limits on what a computer

can do?

6

Abstract Machine ModelsDFA’s

DFA’s model computers with strictly bounded memory.

1

3

2

a

b

b

a

a,b

7

Abstract Machine ModelsDFA’s

Q: What’s the accepted language?

1

3

2

a

b

b

a

a,b

8

Abstract Machine ModelsDFA’s

A: a*b+

1

3

2

a

b

b

a

a,b

9

Abstract Machine ModelsNFA’s

Nondeterminism is a powerful concept. Often 1st view of a problem is nondeterministic.

1

3

2

a

a

b

a

a,b

10

Abstract Machine ModelsNFA’s

Q: What’s the accepted language?

1

3

2

a

a

b

a

a,b

11

Abstract Machine ModelsNFA’s

A: a+b*

1

3

2

a

a

b

a

a,b

12

Abstract Machine ModelsPDA’s

By allowing a pushdown stack, increase flexibility and accept more languages.

1 2

3

a,X

b,X

0

$ $

13

Abstract Machine ModelsPDA’s

Q: What’s the accepted language?

1 2

3

a,X

b,X

0

$ $

14

Abstract Machine ModelsPDA’s

A: {an bn | n 0}

1 2

3

a,X

b,X

0

$ $

15

Abstract Machine ModelsTM’s

By allowing a read-write tape, amazingly get most general possible computer model!

1 2

XR

0

1$,RL acc

34XL

$L

1L

$R

1X,R

1R

XR

XL

5 1L

1|XL

16

Abstract Machine ModelsTM’s

Q: What’s the accepted language?

1 2

XR

0

1$,RL acc

34XL

$L

1L

$R

1X,R

1R

XR

XL

5 1L

1|XL

17

Abstract Machine ModelsTM’s

A: Unary powers of 2.

1 2

XR

0

1$,RL acc

34XL

$L

1L

$R

1X,R

1R

XR

XL

5 1L

1|XL

18

I/O VersionsEach class of languages has its own I/O

version. Regular: Finite State Transducers More powerful models exist (e.g. with

’s)

Context free: (didn’t study any) “Compilers”: Input is a string, output is

a parse-tree (or even executable code)

Turing Machines: I/O TM’s

19

Robust Formal Language Classes

Turns out these models are very robust Many equivalent ways to generate same classes: Regular languages

FA’s, NFA’s, Regular Expressions, Right-Linear Grammars Context Free Languages

PDA’s, Context Free Grammars Recognizable languages –Church-Turing thesis

TM’s, k-tape machines, k-track machines NTM’s, Queue Machines, 2-Stack PDA’s, RAM’s, Unrestricted Grammars

Complexity classes P and NP For NP: Poly. NTM’s, Poly. Verifiers, Poly. Proofs

We learned algorithms for converting between most of the different views

Language classes closed under natural operations.

20

System DesignOften computer system creation involves designing

a formal language to describe system communication. Components receive communication streams and have to effect actions based on these. Computability theory can help drive design at a high level.

EG: Might come up with a communication stream that’s seems like regular language. Could then show that it isn’t using pumping lemma. With this knowledge, final design tweaks original to obtain a regular language and therefore DFA based ultra-fast and super-reliable system components!

21

Negative Examples

As above, it is very important to be able to tell when particular languages cannot be accepted by a certain model of computation. We have several tools at our disposal: Irregularity: pumping lemma (PL) Non-Context-Freeness: CFPL Undecidability: Reductions from undecidable languages Intractability: Poly-time reduction from NP-hard languages

22

System Design

Learned useful concepts that can help modularization when designing systems. Often can express a language as a union, intersection, negation, concatenation or Kleene-* of simpler languages. More complex language may be put together by using simple components along with “off the shelf” reconstruction techniques:

Language Design

Class Negate Concat.

Kleene-*

DFA’sCartesianProduct

Accept Non-

accepts

NFA’s ParallelCartesia

n Product

Serial Loop

PDA’s Parallel Serial Loop

Deciders Run in parallel

Accept Non-

accepts

Break string

up

Recursive

algorithm

Recognizers Run in parallel ” ”

24

Language Class Hierarchy

AllREC = accepted by TM

DEC = decided by TMContext Free

Deterministic Context Free

Regular = accepted by FA’s

Finite languages

25

KnownComplexity Hierarchy

Get the following RAM hierarchy diagram: REC

DECP

TIME(n)

CFLTIME(n3)

REG

EQREX

26

UnknownComplexity Hierarchy

DecidableNP

NP but not NP-hard

P

Finite languages

Does anythingexist here?

27

Conjectured HierarchyInside of DEC most conjecture:

DEC

NP P co-NPNPcomplete

PRIMESAT

28

Follow-ups to Computability

Related Electives: Analysis of Algorithms 4231 (fall) Computational Complexity 4236 (spring) Cryptography –generic course no. 4995

(this Spring with Michael Rabin!!!)

Courses Requiring Computability: Programming Languages and

Translators 4115 (every semester) Compilers 4117 (this Spring with Al

Aho!!!) Portions of several other courses

29

Final Remarks

With the horrors at the beginning of the semester….

30

Final Remarks

Thanks for putting in the effort and

helping make this my best semester

thus far!