1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

40
1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University
  • date post

    15-Jan-2016
  • Category

    Documents

  • view

    220
  • download

    0

Transcript of 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Page 1: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

1

Understanding Problem Hardness: Recent Developments and Directions

Bart Selman Cornell University

Understanding Problem Hardness: Recent Developments and Directions

Bart Selman Cornell University

Page 2: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

2

Introduction & MotivationIntroduction & Motivation

Computational Challenges in Planning, Reasoning,

Learning, and Adaptation.

What are the characteristics of challenging

computational problems?

Page 3: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

A Few Examples

Reasoning

many forms of deductionabduction / diagnosis (e.g. de Kleer 1989)

default reasoning (e.g. Kautz and Selman 1989)

Bayesian inference (e.g. Dagum and Luby 1993)

Planningdomain-dependent and independent (STRIPS)

(e.g. Chapman 1987; Gupta and Nau 1991; Bylander1994)

Learning

neural net “loading” problem (e.g. Blum and Rivest 1989)

Bayesian net learning

decision tree learning

Page 4: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

An abundance of negative complexity results for

many interesting tasks.

Results often apply to very restricted formalisms,

and also to finding approximate solutions.

But worst-case, what about average-case?

Sometimes “surprising” results.

A closer look leads to new insights &

algorithms and solution strategies.

Page 5: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

5

OutlineOutline

A --- “Early’’ results:

phase transitions & computational hardness

B --- Current focus:

--- problem mixtures (tractable / intractable)

--- adding global structure

C --- Future directions and prospects

--- modeling resource constraints

--- adaptive computing

--- deeper theoretical understanding

Page 6: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

6

A. “Early” Results (‘90-’95)

Page 7: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Example Domain: Satisfiability

SAT: Given a formula in propositional calculus, is there an assignment to its variables making it true?

We consider clausal form, e.g.:a b cb db c

e

The canonical NP-complete problem.

(“exponential search space”)

Page 8: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Variable Clause Size Model

Polynominal average time in regions: Ia Ð Purdom 1987 - backtracking Ib Ð Iwama 1989 - counting alg. Ic Ð Brown and Purdom 1985 - pure literal rule II Ð Franco 1991 III Ð Franco 1994 Open: region IV

Ratio of Clauses-to-Variables

m-1/2

m-2 m-1

m1/2

1Avera

ge C

lause L

ength

1

Ia

Ib

Ic

II

IV

III?

Page 9: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Generating Hard Random Formulas

Key: Use fixed-clause-length model.(Mitchell, Selman, and Levesque 1992; Kirkpatrick and

Selman 1994)

Critical parameter: ratio of the number of clauses to the number of variables.

Hardest 3SAT problems at ratio = 4.25

Page 10: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Hardness of 3SAT

02 3 4 5

Ratio of Clauses-to-Variables

6 7 8

1000

3000

DP

Calls

2000

4000

50 var 40 var 20 var

Page 11: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Intuition

At low ratios:few clauses (constraints)many assignmentseasily found

At high ratios:many clausesinconsistencies easily detected

Page 12: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

The 4.3 Point

0.02 3 4 5

Ratio of Clauses-to-Variables

6 7 8

0.2

0.6

Pro

bab

ilit

yD

P C

alls

0.4

50 var 40 var 20 var

50% sat

Mitchell, Selman, and Levesque 1991

0.8

1.0

0

1000

3000

2000

4000

Page 13: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

13

Phase transition 2-, 3-, 4-, 5-, and 6-SATPhase transition 2-, 3-, 4-, 5-, and 6-SAT

Page 14: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Theoretical Status Of Threshold

Very challenging problem ...

Current status:

3SAT threshold lies between 3.003 and 4.6. (Motwani et al. 1994; Broder et al. 1992;

Frieze and Suen 1996; Dubois 1990, 1997;

Kirousis et al. 1995; Friedgut 1997;

Archlioptas et al. 1999 / related work:

Beame, Karp, Pitassi, and Saks 1998;

Bollobas, Borgs, Chayes, Han Kim, and

Wilson 1999)

Page 15: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Phase transition and combinatorial problems is an

active research area with fruitful interactions

between computer science, physics (approaches

from statistical mechanics), and mathematics

(combinatorics / random structures).

Also, a close interaction between experimental and

theoretical work. (With experimental findings quite often

confirmed by formal analysis within months to a few years.)

Finally, relevance to applications via algorithmic

advances and notion of “critically constrained

problems”.

Page 16: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Consequences for Algorithm Design

Phase transition work instances led to

improvements in algorithms:

--- local search methods (e.g., GSAT / Walksat)

(Selman et al. 1992; 1996; Min Li 1996; Hoos 1998, etc.)

--- backtrack-style methods (Davis-Putnam and

variants / complete)

(Crawford 1993; Dubois 1994; Bayardo 1997; Zane 1998, etc.)

Page 17: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

17

ProgressProgress

Propositional reasoning and search (SAT):

1990: 100 variables / 200 clauses (constraints)

1998: 10,000 - 100,000 variables / 10^6 clauses

Novel applications:

e.g. in planning (Kautz & Selman),

program debugging (Jackson),

protocol verification (Clarke), and

machine learning (Resende).

Page 18: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

B. Current Focus

--- mixtures of problem classes, e.g., 2-SAT

and 3-SAT (“moving between P and NP”)

the 2+p-SAT model

--- structured instances

perturbed quasi-group completion problems

Page 19: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Focus --- 1) mixtures: 2+p-SAT problem

mixture of binary and ternary clauses

p = fraction ternary

p = 0.0 --- 2-SAT / p = 1.0 --- 3-SAT

What happens in-between?

(Monasson, Zecchina, Kirkpatrick, Selman, and Troyansky,

Nature, to appear)

Page 20: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

20

Phase Transition for 2+p-SAT Phase Transition for 2+p-SAT

Page 21: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

21

Location ThresholdLocation Threshold

Page 22: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

22

Computational CostComputational Cost

Page 23: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

23

Results for 2+p-SATResults for 2+p-SAT

p < ~ 0.41 --- model essentially behaves as 2-SAT

search proc. “sees” only binary constraints

smooth, continuous phase transition

p > ~ 0.41 --- behaves as 3-SAT (exponential scaling)

abrupt, discontinuous scaling

Many new, rigorous results (including scaling) by

Achlioptas, Bollobas, Borgs, Chayes, Han Kim,

and Wilson. (Next talk.)

Page 24: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Consequences for Algorithm Design

1) Strategies that exploit tractable substructure

with propagation are most effective.

(consistent with the best empirically discovered

methods)

2) In addition, use early branching on critically

constrained variables.

(the “backbone variables” / suggests use of

clustering and statistical learning methods)

(Boyan and Moore 1998)

Page 25: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

25

Proposal: study the influence of globalProposal: study the influence of global

structure on problem hardness.structure on problem hardness.

Focus --- 2) StructureFocus --- 2) Structure

(Gomes and Selman 1997; 1998)

Page 26: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

26

Defn.: a pair (Q, *) where Q is a set, and * is a binary

operation on Q such that

a * x = b ; y * a = b

are uniquely solvable for every pair of elements a,b in Q.

The multiplication table of its binary operation defines a

latin square (i.e., each element of Q appears exactly once

in each row/column).

Example:Quasigroup of order 4

QuasigroupsQuasigroups

Page 27: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

27

Given a partial latin square, can it be completed?

Example:

Quasigroup Completion Problem (QCP)

Quasigroup Completion Problem (QCP)

Page 28: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

28

Quasigroup Completion Problem A Framework for Studying SearchQuasigroup Completion Problem

A Framework for Studying Search

NP-Complete (Colbourn 1983, 1984; Anderson 1985).

Has a regular global structure not found in

random instances.

Leads to interesting search problems when

structure is perturbed.

similar to e.g. structure found in the channel assignment problem

for cellular networks

Page 29: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

29

Computational CostComputational Cost

Page 30: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

30

Page 31: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Consequences for Algorithm Design

On these structured problems, backtrack

search methods show so-called

heavy-tailed probability distributions.

(Gomes, Selman & Crato 1997, 1998).

Both very short and very long runs occur

much more frequent than one would expect.

Page 32: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

32

Standard Distribution

Page 33: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

33

Heavy Tailed Cost DistributionHeavy Tailed Cost Distribution

0.1

1

1 10 100 1000 10000 100000

log( Backtracks )

log

( 1

- F

(x)

)

Page 34: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

34

Fringe of Search TreeFringe of Search Tree

Page 35: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Algorithmic Strategy:

Rapid Random Restarts.

Order of magnitude speedup.

(Gomes et al. 1998; 1999)

Related:

. Algorithm portfolios (Huberman 1998; Gomes 1998)

. Universal strategies

(Ertel and Luby 1993; Alt et al. 1996)

Page 36: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

36

Rapid Restarts --- PlanningRapid Restarts --- Planning

1000

10000

100000

1000000

1 10 100 1000 10000 100000 1000000

log( cutoff )

log

( b

ackt

rack

s )

Page 37: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

37

Portfolio for heavy-tailed search procedures (2-20 processors)

Portfolio for heavy-tailed search procedures (2-20 processors)

Page 38: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

38

C. Future directions and prospectsC. Future directions and prospects

Modeling resource constraints &

user requirements / utility

should be possible to identify optimal

restart strategies, possibly adaptive

--- may need way of “measuring progress” (Horvitz and Klein 1995; Gomes and Selman 1999)

Page 39: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

39

Adaptive Computing

combine statistical learning methods with

combinatorial search techniques.

first success: STAGE system for local search.

(Boyan and Moore 1998)

extension: train a planner on small instances

(Selman, Kautz, Huang 1999)

Deeper theoretical understanding with continued interactions with experiments

and applications

Page 40: 1 Understanding Problem Hardness: Recent Developments and Directions Bart Selman Cornell University.

Summary

During the past few years, we have obtained a much

better understanding of the nature of

computationally hard problems.

Rich interactions between physics, computer

science and mathematics, and between theory,

experiments, and applications.

Clear algorithmic progress with room for future improvements (possibly another level of scaling:

10^6 Boolean variables, 10^8 constraints. Further

applications.)