adaptive bisection of numerical csps
play

Adaptive Bisection of Numerical CSPs Laurent Granvilliers Univ. - PowerPoint PPT Presentation

Adaptive Bisection of Numerical CSPs Laurent Granvilliers Univ. Nantes, Lab. Computer Science, France L. Granvilliers (Nantes) Adaptive Bisection CP 2012 1 / 10 Bisection Algorithm Goal : Solving numerical CSPs using interval computations. x


  1. Adaptive Bisection of Numerical CSPs Laurent Granvilliers Univ. Nantes, Lab. Computer Science, France L. Granvilliers (Nantes) Adaptive Bisection CP 2012 1 / 10

  2. Bisection Algorithm Goal : Solving numerical CSPs using interval computations. x 2 x 1 ( x 2 1 + x 2 2 ) = 6 x 2  2  ( x 1 − 0 . 25) 2 + x 2 2 = 4 ( x 1 , x 2 ) ∈ x 1 × x 2  x 1 x = x 1 × x 2 L. Granvilliers (Nantes) Adaptive Bisection CP 2012 2 / 10

  3. Bisection Algorithm Goal : Solving numerical CSPs using interval computations. x 2 x 1 ( x 2 1 + x 2 2 ) = 6 x 2  2  ( x 1 − 0 . 25) 2 + x 2 2 = 4 ( x 1 , x 2 ) ∈ x 1 × x 2  • Contract x → x ′ x 1 x ′ L. Granvilliers (Nantes) Adaptive Bisection CP 2012 2 / 10

  4. Bisection Algorithm Goal : Solving numerical CSPs using interval computations. x 2 x 1 ( x 2 1 + x 2 2 ) = 6 x 2  2  ( x 1 − 0 . 25) 2 + x 2 2 = 4 ( x 1 , x 2 ) ∈ x 1 × x 2  • Contract x → x ′ x 1 • Select x 2 • Bisect x ′ along x 2 x ′ L. Granvilliers (Nantes) Adaptive Bisection CP 2012 2 / 10

  5. Bisection Algorithm Goal : Solving numerical CSPs using interval computations. x 2 x 1 ( x 2 1 + x 2 2 ) = 6 x 2  2  ( x 1 − 0 . 25) 2 + x 2 2 = 4 ( x 1 , x 2 ) ∈ x 1 × x 2  • Contract x → x ′ x 1 • Select x 2 • Bisect x ′ along x 2 • Iterate and stop on ǫ -boxes L. Granvilliers (Nantes) Adaptive Bisection CP 2012 2 / 10

  6. MaxDom Strategy Let f ( x 1 , x 2 ) = x 1 ( x 2 1 + x 2 2 ) − 6 x 2 2 be the function defining the cissoid. Let x = [2 , 4] × [0 , 1] and evaluate the natural extension of f : f ( x ) = [2 , 4] ([2 , 4] 2 + [0 , 1] 2 ) − 6 [0 , 1] 2 = [2 , 68] L. Granvilliers (Nantes) Adaptive Bisection CP 2012 3 / 10

  7. MaxDom Strategy Let f ( x 1 , x 2 ) = x 1 ( x 2 1 + x 2 2 ) − 6 x 2 2 be the function defining the cissoid. Let x = [2 , 4] × [0 , 1] and evaluate the natural extension of f : f ( x ) = [2 , 4] ([2 , 4] 2 + [0 , 1] 2 ) − 6 [0 , 1] 2 = [2 , 68] Bisect x 1 Bisect x 2 f ([2 , 3] , x 2 ) = [2 , 30] f ( x 1 , [0 , 0 . 5]) = [6 . 5 , 65] f ([3 , 4] , x 2 ) = [21 , 68] f ( x 1 , [0 . 5 , 1]) = [2 . 5 , 66 . 5] L. Granvilliers (Nantes) Adaptive Bisection CP 2012 3 / 10

  8. MaxDom Strategy Let f ( x 1 , x 2 ) = x 1 ( x 2 1 + x 2 2 ) − 6 x 2 2 be the function defining the cissoid. Let x = [2 , 4] × [0 , 1] and evaluate the natural extension of f : f ( x ) = [2 , 4] ([2 , 4] 2 + [0 , 1] 2 ) − 6 [0 , 1] 2 = [2 , 68] Bisect x 1 Bisect x 2 f ([2 , 3] , x 2 ) = [2 , 30] f ( x 1 , [0 , 0 . 5]) = [6 . 5 , 65] f ([3 , 4] , x 2 ) = [21 , 68] f ( x 1 , [0 . 5 , 1]) = [2 . 5 , 66 . 5] MaxDom : bisecting the largest domain reduces overestimation in interval computations = ⇒ more contraction using local consistency L. Granvilliers (Nantes) Adaptive Bisection CP 2012 3 / 10

  9. MaxSmear Strategy Let x = [0 . 5 , 1] × [0 . 5 , 1] and evaluate the derivatives of f : ∇ f ( x ) = ([1 , 4] , [ − 11 . 5 , − 4]) L. Granvilliers (Nantes) Adaptive Bisection CP 2012 4 / 10

  10. MaxSmear Strategy Let x = [0 . 5 , 1] × [0 . 5 , 1] and evaluate the derivatives of f : ∇ f ( x ) = ([1 , 4] , [ − 11 . 5 , − 4]) Let c be the midpoint of x and evaluate the mean value extension of f : f ( x , c ) = f ( c ) + ∇ f ( x )( x − c ) = [ − 6 . 5 , 1 . 4] L. Granvilliers (Nantes) Adaptive Bisection CP 2012 4 / 10

  11. MaxSmear Strategy Let x = [0 . 5 , 1] × [0 . 5 , 1] and evaluate the derivatives of f : ∇ f ( x ) = ([1 , 4] , [ − 11 . 5 , − 4]) Let c be the midpoint of x and evaluate the mean value extension of f : f ( x , c ) = f ( c ) + ∇ f ( x )( x − c ) = [ − 6 . 5 , 1 . 4] Bisect x 1 Bisect x 2 f ([0 . 5 , 0 . 75] , x 2 ) = [ − 6 . 0 , 0 . 5] f ( x 1 , [0 . 75 , 1]) = [ − 6 . 0 , − 1 . 2] f ([0 . 75 , 1 . ] , x 2 ) = [ − 5 . 6 , 1 . 1] f ( x 1 , [0 . 5 , 0 . 75]) = [ − 3 . 6 , 0 . 4] L. Granvilliers (Nantes) Adaptive Bisection CP 2012 4 / 10

  12. MaxSmear Strategy Let x = [0 . 5 , 1] × [0 . 5 , 1] and evaluate the derivatives of f : ∇ f ( x ) = ([1 , 4] , [ − 11 . 5 , − 4]) Let c be the midpoint of x and evaluate the mean value extension of f : f ( x , c ) = f ( c ) + ∇ f ( x )( x − c ) = [ − 6 . 5 , 1 . 4] Bisect x 1 Bisect x 2 f ([0 . 5 , 0 . 75] , x 2 ) = [ − 6 . 0 , 0 . 5] f ( x 1 , [0 . 75 , 1]) = [ − 6 . 0 , − 1 . 2] f ([0 . 75 , 1 . ] , x 2 ) = [ − 5 . 6 , 1 . 1] f ( x 1 , [0 . 5 , 0 . 75]) = [ − 3 . 6 , 0 . 4] MaxSmear : selecting the variable having the maximum smear value produces tighter linear relaxations = ⇒ Newton operator stronger L. Granvilliers (Nantes) Adaptive Bisection CP 2012 4 / 10

  13. RoundRobin Strategy • Total order x i 1 < x i 2 < · · · < x i n – x i < x j if x i occurs more than x j – x i < x j if x i occurs in more constraints than x j – . . . L. Granvilliers (Nantes) Adaptive Bisection CP 2012 5 / 10

  14. RoundRobin Strategy • Total order x i 1 < x i 2 < · · · < x i n – x i < x j if x i occurs more than x j – x i < x j if x i occurs in more constraints than x j – . . . • RoundRobin : select x i 1 , x i 2 , . . . , x i n , x i 1 , x i 2 , . . . x i n , x i 1 , . . . – every domain is regularly bisected – fair strategy L. Granvilliers (Nantes) Adaptive Bisection CP 2012 5 / 10

  15. New Adaptive Strategy • Motivation – robust smear strategy – efficient fair strategy L. Granvilliers (Nantes) Adaptive Bisection CP 2012 6 / 10

  16. New Adaptive Strategy • Motivation – robust smear strategy – efficient fair strategy • Adaptive bisection strategy – diversification ( ≈ RoundRobin) in the early steps of the algorithm – intensification ( ≈ MaxSmear) in the vicinity of solutions L. Granvilliers (Nantes) Adaptive Bisection CP 2012 6 / 10

  17. New Adaptive Strategy • Motivation – robust smear strategy – efficient fair strategy • Adaptive bisection strategy – diversification ( ≈ RoundRobin) in the early steps of the algorithm – intensification ( ≈ MaxSmear) in the vicinity of solutions • GRASP (Feo & Resende 1995) – adaptive search procedure for derivative-free optimization – greedy + randomized L. Granvilliers (Nantes) Adaptive Bisection CP 2012 6 / 10

  18. New Algorithm • Initialization : for every variable ( i = 1 , . . . , n ) – s i ≥ 0 : smear value of x i in x – n i ≥ 0 : number of times x i has been selected – [ s min , s max ] : range of smear values L. Granvilliers (Nantes) Adaptive Bisection CP 2012 7 / 10

  19. New Algorithm • Initialization : for every variable ( i = 1 , . . . , n ) – s i ≥ 0 : smear value of x i in x – n i ≥ 0 : number of times x i has been selected – [ s min , s max ] : range of smear values • Subset of the best variables : let S ∈ [ s min , s max ] and mark every variable x j s.t. s j ≥ S ∧ width( x j ) > ǫ L. Granvilliers (Nantes) Adaptive Bisection CP 2012 7 / 10

  20. New Algorithm • Initialization : for every variable ( i = 1 , . . . , n ) – s i ≥ 0 : smear value of x i in x – n i ≥ 0 : number of times x i has been selected – [ s min , s max ] : range of smear values • Subset of the best variables : let S ∈ [ s min , s max ] and mark every variable x j s.t. s j ≥ S ∧ width( x j ) > ǫ • Fair choice : select a marked variable x k s.t. n k = min { n i : 1 ≤ i ≤ n ∧ x i marked } L. Granvilliers (Nantes) Adaptive Bisection CP 2012 7 / 10

  21. Adaptive Behaviour • Threshold on smear values S = s min + α ( s max − s min ) , 0 ≤ α ≤ 1 – α = 1 = ⇒ greedy – α = 0 = ⇒ fair L. Granvilliers (Nantes) Adaptive Bisection CP 2012 8 / 10

  22. Adaptive Behaviour • Threshold on smear values S = s min + α ( s max − s min ) , 0 ≤ α ≤ 1 – α = 1 = ⇒ greedy – α = 0 = ⇒ fair • Adaptive behaviour 1 α = 1 + βσ , β > 0 – σ : standard deviation of the smear values – σ → 0 = ⇒ α → 1 – σ → ∞ = ⇒ α → 0 L. Granvilliers (Nantes) Adaptive Bisection CP 2012 8 / 10

  23. Experimental Results Problem MaxDom RoundRobin MaxSmear Adaptive n Celestial 3 3 361 2 599 4 258 3 074 Combu. 10 559 463 1 303 509 Neuro 6 1 288 779 7 223 7 215 6 367 Trigo1 10 1 258 2 086 2 071 1 244 Broyden3 20 24 723 23 23 Geineg 6 8 546 8 116 1 707 2 400 Kapur 5 2 791 1 651 231 323 Nbody 8 1 976 2 049 1 532 1 765 Brown 10 12 443 5 850 6 827 4 996 Eco 6 1 108 1 530 1 063 1 023 Nauheim 8 1 204 722 816 708 Transistor 9 112 795 121 765 83 711 42 085 L. Granvilliers (Nantes) Adaptive Bisection CP 2012 9 / 10

  24. Experimental Results Problem MaxDom RoundRobin MaxSmear Adaptive n Celestial 3 3 361 2 599 4 258 3 074 Combu. 10 559 463 1 303 509 Neuro 6 1 288 779 7 223 7 215 6 367 Trigo1 10 1 258 2 086 2 071 1 244 Broyden3 20 24 723 23 23 Geineg 6 8 546 8 116 1 707 2 400 Kapur 5 2 791 1 651 231 323 Nbody 8 1 976 2 049 1 532 1 765 Brown 10 12 443 5 850 6 827 4 996 Eco 6 1 108 1 530 1 063 1 023 Nauheim 8 1 204 722 816 708 Transistor 9 112 795 121 765 83 711 42 085 L. Granvilliers (Nantes) Adaptive Bisection CP 2012 9 / 10

Recommend


More recommend