Consistency for Quantified Constraint Satisfaction Problems Peter - - PowerPoint PPT Presentation

consistency for quantified constraint satisfaction
SMART_READER_LITE
LIVE PREVIEW

Consistency for Quantified Constraint Satisfaction Problems Peter - - PowerPoint PPT Presentation

Consistency for Quantified Constraint Satisfaction Problems Peter Nightingale Talk structure Finite domain QCSP Connect-4 Consistency notions WQGAC WQGAC-Schema Comparing consistencies Summary Finite domain QCSP


slide-1
SLIDE 1

Consistency for Quantified Constraint Satisfaction Problems Peter Nightingale

slide-2
SLIDE 2

Talk structure

  • Finite domain QCSP – Connect-4
  • Consistency notions
  • WQGAC
  • WQGAC-Schema
  • Comparing consistencies
  • Summary
slide-3
SLIDE 3

Finite domain QCSP

  • Connect-4 endgame

1 2 3 4 5 6 7

∃red1∀black1∃red2∀ black2∃red3: redwinsred1,black1,red2,black2,red3

slide-4
SLIDE 4

red1 black1 red2 red2 2 2 1,3..7 black2 SAT 5 2 red3 red3 4 1..3,5..7 SAT SAT 5 4

Finite domain QCSP

  • Example strategy

1 2 3 4 5 6 7

slide-5
SLIDE 5

red1 black1 red2 red2 2 2 1,3..7 black2 SAT 5 2 red3 red3 4 1..3,5..7 SAT SAT 5 4

Finite domain QCSP

  • Example strategy

1 2 3 4 5 6 7

slide-6
SLIDE 6

red1 black1 red2 red2 2 1,3..7 black2 SAT 5 2 red3 red3 4 1..3,5..7 SAT SAT 5 4

Finite domain QCSP

  • Example strategy

1 2 3 4 5 6 7

2

slide-7
SLIDE 7

red1 black1 red2 red2 2 1,3..7 black2 SAT 5 2 red3 red3 4 1..3,5..7 SAT SAT 4 4

Finite domain QCSP

  • Example strategy

1 2 3 4 5 6 7

2

slide-8
SLIDE 8

red1 black1 red2 red2 2 1,3..7 black2 SAT 5 2 red3 red3 4 1..3,5..7 SAT SAT 4 4

Finite domain QCSP

  • Example strategy

1 2 3 4 5 6 7

2

slide-9
SLIDE 9

red1 black1 red2 red2 2 1,3..7 black2 SAT 5 2 red3 red3 4 1..3,5..7 SAT SAT 4 4

Finite domain QCSP

  • Example strategy

1 2 3 4 5 6 7

2

slide-10
SLIDE 10

Talk structure

  • Finite domain QCSP – Connect-4
  • Consistency notions
  • WQGAC
  • WQGAC-Schema
  • Comparing consistencies
  • Summary
slide-11
SLIDE 11

Consistency notions

  • Hasse diagram
  • Ordered by strength

– Then constraint arity

Ternary Boolean constraints

Bordeaux and Monfroy

QAC

Stergiou and Mamoulis

WQGAC

(this work)

Ternary interval constraints

Bordeaux and Monfroy

AC GAC Local inconsistency

Bordeaux, Cadoli and Mancini

slide-12
SLIDE 12

Talk structure

  • Finite domain QCSP – Connect-4
  • Consistency notions
  • WQGAC
  • WQGAC-Schema
  • Comparing consistencies
  • Summary
slide-13
SLIDE 13

WQGAC

  • With GAC each value has a supporting

tuple

  • With WQGAC each value has a supporting

tuple for each combination of values of inner universals

∃a ∀ b∃c:a⇔b∧c

a b c 0 0 0 0 0 1 0 1 0 1 1 1 Supporting a=0:

slide-14
SLIDE 14

WQGAC

  • With GAC each value has a supporting

tuple

  • With WQGAC each value has a supporting

tuple for each combination of values of inner universals ∃a∀ b∃c:a⇔b∧c a b c 0 0 0 0 0 1 0 1 0 1 1 1 Supporting a=1:

slide-15
SLIDE 15

WQGAC-Schema

  • Based on GAC-Schema (Bessière and

Régin)

  • Time: O(n2dn)
  • Space: O(n2du+1)
  • Generalization of GAC-Schema
  • Multidirectional
slide-16
SLIDE 16

Talk structure

  • Finite domain QCSP – Connect-4
  • Consistency notions
  • WQGAC
  • WQGAC-Schema
  • Comparing consistencies
  • Summary
slide-17
SLIDE 17

Comparing consistencies

Consistency QAC on the hidden variable encoding GAC WQGAC B,C & M inconsistency Inference none none 1,3,5..7 pruned from grey1 1,3,5..7 pruned from grey1 1,3,6,7 pruned from grey2 1,3,7 pruned from grey3 Resources used 0.046s, checked 15.2% of all 75 tuples.

slide-18
SLIDE 18

Comparing consistencies

  • WQGAC weak

– For each value, set of supporting tuples – May not be part of one strategy

a=0 supported by: a b c 0 0 0 0 1 1 1 0 1 1 1 0

∀ a∃b ∀c∈{0,1}

Value of b is different

slide-19
SLIDE 19

Summary

  • Reasonably powerful algorithm for local

reasoning in finite domain QCSP

  • Future work

– Tuple/tree mismatch – Different support structure

slide-20
SLIDE 20

Thank you

  • Any questions?