Search Combinators Tom Schrijvers with Guido Tack, Pieter Wuille, - - PowerPoint PPT Presentation

▶
search combinators
SMART_READER_LITE
LIVE PREVIEW

Search Combinators Tom Schrijvers with Guido Tack, Pieter Wuille, - - PowerPoint PPT Presentation

Dec 10, 2022 181 likes •562 views

Search Combinators Tom Schrijvers with Guido Tack, Pieter Wuille, Horst Samulowitz, Peter Stuckey Search heuristics are crucial. Support for Search? General Purpose Solver-Provided Programming Language Options everthing is possible,

slide-1
SLIDE 1

Search Combinators

Tom Schrijvers

with Guido Tack, Pieter Wuille, Horst Samulowitz, Peter Stuckey

slide-2
SLIDE 2

Search heuristics are crucial.

slide-3
SLIDE 3

Support for Search?

General Purpose Programming Language “everthing is possible, nothing is easy” Solver-Provided Options “everything is easy, nothing is possible”

slide-4
SLIDE 4

State-of-the-Art Modularity: Prolog

✓ label1(Vars1) , label2(Vars2) ✓ label1(Vars1) ; label2(Vars2) ✓ once(label1(Vars1)) ✓ once((label1(Vars1),label2(Vars2))

slide-5
SLIDE 5

Lack of Modularity

✘ lds(label1(Vars1))

slide-6
SLIDE 6

Lack of Modularity

✘ lds(label1(Vars1)) ✘ lds((label1(Vars1) , label2(Vars2)))

slide-7
SLIDE 7

Lack of Modularity

✘ lds(label1(Vars1)) ✘ lds((label1(Vars1) , label2(Vars2))) ✓ lds_label1(Vars1) ✓ lds_label1_label2(Vars1,Vars2)

slide-8
SLIDE 8

Can we do better?

✓Lots of expressivity and flexibility ✓Lots of productivity through high-level

specifications

✓Modular reuse of search specifications

slide-9
SLIDE 9

Yes: Search Combinators

“Everything is possible and easy” High-level modular building blocks

slide-10
SLIDE 10

Base Search

s ≡ int_search(vars,var_sel,val_sel)

  • provided by the solver
  • augment with combinators
slide-11
SLIDE 11

Combinators

prune let(v,e,s) assign(v,e) post(c,s) if(c,s1,s2) and([s1, s2,..., sn])

  • r([s1, s2,..., sn])

portfolio([s1, s2,..., sn]) restart(c,s)

slide-12
SLIDE 12

Statistics Combinators

depth(v,s) discrepancy(v,s) nodes(v,s) failures(v,s) time(v,s)

slide-13
SLIDE 13

Statistics Combinators

depth(v,s) discrepancy(v,s) nodes(v,s) failures(v,s) time(v,s) if(v < 5,depth(v,s1),s2)

slide-14
SLIDE 14

Statistics Combinators

depth(v,s) discrepancy(v,s) nodes(v,s) failures(v,s) time(v,s) if(v < 5,depth(v,s1),s2) if(depth < 5,s1,s2)

slide-15
SLIDE 15

Reusable Abstractions

limit(c,s) ≡ if(c,s,prune) for(v,l,u,s) ≡ ... lds(s) ≡ for(n,0,∞, limit(discrepancy ≤ n,s) )

slide-16
SLIDE 16

Reuse Examples

✓lds(int_search(vars,...)) ✓lds(and([int_search(vars1,...)

,int_search(vars2,...)])

✓lds(or([int_search(vars1,...)

,int_search(vars2,...)]) ...

slide-17
SLIDE 17

More Abstractions

bab(obj,s) restart_bab(obj,s) dicho(obj,s,lb,ub) id(s) hot_start(c,s1,s2) see paper

slide-18
SLIDE 18

Radiotherapy Planning

bab(k, and([int_search(N,...)] ++ [once(int_search(rowi,...)) | i in 1..n ] ) )

slide-19
SLIDE 19

Modular Syntax vs. Semantics Modular

slide-20
SLIDE 20

Modular Semantics

traditional modularity:

  • each combinator is an

atomic unit

  • minimal interaction

search tree node

slide-21
SLIDE 21

Modular Semantics

traditional modularity:

  • each combinator is an

atomic unit

  • minimal interaction

search tree node

slide-22
SLIDE 22

Modular Semantics

  • cross-cutting behavior
  • highly entangled in

monolithic code

➡ “Aspect-Oriented Programming”

slide-23
SLIDE 23

Functional Mixin Inheritance

  • disciplined form of AOP

★ meaningful and pre-defined set of

interaction points

  • no AOP system needed
slide-24
SLIDE 24

Mixin-based Interaction

success failure enter(n) more for every child c push(c) next(n',n)

slide-25
SLIDE 25

Mixin-based Interaction

Details: see paper

success failure enter(n) more for every child c push(c) next(n',n)

slide-26
SLIDE 26

Implementations

Compacte Lus

DSL Haskell C++ Scala

Compact Loop

slide-27
SLIDE 27

Implementations

Zoekheuristiek Compacte Lus

DSL Haskell C++ Scala

Compact Loop Search Spec

slide-28
SLIDE 28

Implementations

Zoekheuristiek Objectcompositie Compacte Lus

DSL Haskell C++ Scala

Compact Loop Search Spec Objects

Interpreted

slide-29
SLIDE 29

Implementations

Zoekheuristiek Objectcompositie Compacte Lus Codegenerators

DSL Haskell C++ Scala

Compact Loop Search Spec Objects

Code Generators

Interpreted Compiled

slide-30
SLIDE 30

Combinator Overhead?

portfolio prune base search without propagation

n

Worst-case Scenario

100,00% 150,00% 200,00% 250,00% 300,00%

1 2 5 10 20

runtime

#combinators

compiled interpreted

slide-31
SLIDE 31

80,00% 100,00% 120,00% 140,00% 160,00% 180,00%

G

  • l
  • m

b 1 G

  • l
  • m

b 1 1 G

  • l
  • m

b 1 2 R a d i

  • t

h e r a p y 1 R a d i

  • t

h e r a p y 2 R a d i

  • t

h e r a p y 3 R a d i

  • t

h e r a p y 4 R a d i

  • t

h e r a p y 5 J

  • b
  • s

h

  • p

G 2 J

  • b
  • s

h

  • p

H 5 J

  • b
  • s

h

  • p

H 3 J

  • b
  • s

h

  • p

A B Z 1

  • 5

J

  • b
  • s

h

  • p

m t 1

In Practice

propagation dwarfs combinator overhead

Gecode Interpreted Compiled

slide-32
SLIDE 32

Summary

high-level modular modeling of search low-level modular implementation competitive performance compared to hand-coded algorithm

slide-33
SLIDE 33

Future Work

  • Optimizations
  • MiniZinc integration
  • Combinators for parallel search
  • Extend other solving technology (e.g., LP)

➡ Combinators for hybrid search

slide-34
SLIDE 34

Thank You!

slide-35
SLIDE 35
slide-36
SLIDE 36

Combinator Overhead

100,00% 150,00% 200,00% 250,00% 300,00%

1 2 5 10 20

runtime

#combinators

compiled interpreted