a subexponential lower bound for zadeh s pivoting rule
play

A subexponential lower bound for Zadehs pivoting rule for solving - PowerPoint PPT Presentation

Jun 23, 2023 •140 likes •1.72k views

A subexponential lower bound for Zadehs pivoting rule for solving linear programs and games Oliver Friedmann Department of Computer Science, Ludwig-Maximilians-Universit at Munich, Germany. Oliver Friedmann (LMU) Zadeh Lower Bound 1

  1. Games and Policy Iteration Policies and corresponding values A policy π is a choice of an action from each state. The value val π ( i ) of a state i ∈ S for 6 t a policy π , is the expected sum of rewards obtained when moving 1 according to π , starting from i . 3 2 3 2 0 An action is an improving switch w.r.t. π if it improves the values. -4 0 -1 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 9

  2. Games and Policy Iteration Policies and corresponding values A policy π is a choice of an action from each state. The value val π ( i ) of a state i ∈ S for 6 t a policy π , is the expected sum of rewards obtained when moving 1 according to π , starting from i . 3 2 3 2 0 An action is an improving switch w.r.t. π if it improves the values. It suffices to check whether an action -4 0 is improving for one step w.r.t. the current values. -1 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 9

  3. Games and Policy Iteration Policies and corresponding values A policy π is a choice of an action from each state. The value val π ( i ) of a state i ∈ S for 6 t a policy π , is the expected sum of rewards obtained when moving 1 according to π , starting from i . 3 2 3 6 6 An action is an improving switch w.r.t. π if it improves the values. It suffices to check whether an action -4 0 is improving for one step w.r.t. the current values. -1 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 9

  4. Games and Policy Iteration Policies and corresponding values A policy π is a choice of an action from each state. The value val π ( i ) of a state i ∈ S for 6 t a policy π , is the expected sum of rewards obtained when moving 1 according to π , starting from i . 3 2 3 6 6 An action is an improving switch w.r.t. π if it improves the values. It suffices to check whether an action -4 2 is improving for one step w.r.t. the current values. 1 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 9

  5. Games and Policy Iteration Policies and corresponding values A policy π is a choice of an action from each state. The value val π ( i ) of a state i ∈ S for 6 t a policy π , is the expected sum of rewards obtained when moving 1 according to π , starting from i . 3 2 3 6 6 An action is an improving switch w.r.t. π if it improves the values. It suffices to check whether an action -4 2 is improving for one step w.r.t. the current values. A policy π ∗ is optimal iff there are no 2 improving switches. Optimal policies -1 simultaneously maximize the values of all states. Oliver Friedmann (LMU) Zadeh Lower Bound 9

  6. Games and Policy Iteration MDPs and linear programming No improving switches for optimal policy π ∗ : � ∀ i ∈ S : val π ∗ ( i ) = max a ∈ A i r a + p a,j val π ∗ ( j ) j ∈ S where A i is the set of actions from state i , r a is the expected reward of using action a , and p a,j is the probability of moving to state j when using action a . Oliver Friedmann (LMU) Zadeh Lower Bound 10

  7. Games and Policy Iteration MDPs and linear programming No improving switches for optimal policy π ∗ : � ∀ i ∈ S : val π ∗ ( i ) = max a ∈ A i r a + p a,j val π ∗ ( j ) j ∈ S where A i is the set of actions from state i , r a is the expected reward of using action a , and p a,j is the probability of moving to state j when using action a . This can be used to formulate an LP for solving the MDP: � minimize v i i ∈ S � s.t. ∀ i ∈ S ∀ a ∈ A i : v i ≥ r a + p a,j v j j ∈ S Oliver Friedmann (LMU) Zadeh Lower Bound 10

  8. Games and Policy Iteration Primal and dual LPs for MDPs � minimize v i i ∈ S � s.t. ∀ i ∈ S ∀ a ∈ A i : v i ≥ r a + p a,j v j j ∈ S � � maximize r a x a i ∈ S a ∈ A i � � � s.t. ∀ i ∈ S : x a = 1 + p a,i x a a ∈ A i j ∈ S a ∈ A j Oliver Friedmann (LMU) Zadeh Lower Bound 11

  9. Games and Policy Iteration Primal and dual LPs for MDPs Flow conservation: � minimize v i i ∈ S x 1 = 1 x 2 = 6 � s.t. ∀ i ∈ S ∀ a ∈ A i : v i ≥ r a + p a,j v j j ∈ S � � maximize r a x a x 3 = 4 x 4 = 2 i ∈ S a ∈ A i � � � s.t. ∀ i ∈ S : x a = 1 + p a,i x a x 1 + x 2 = 1 + x 3 + x 4 a ∈ A i j ∈ S a ∈ A j Oliver Friedmann (LMU) Zadeh Lower Bound 11

  10. Games and Policy Iteration Primal and dual LPs for MDPs Flow conservation: � minimize v i i ∈ S x 1 = 7 x 2 = 0 � s.t. ∀ i ∈ S ∀ a ∈ A i : v i ≥ r a + p a,j v j j ∈ S � � maximize r a x a x 3 = 4 x 4 = 2 i ∈ S a ∈ A i � � � s.t. ∀ i ∈ S : x a = 1 + p a,i x a x 1 + x 2 = 1 + x 3 + x 4 a ∈ A i j ∈ S a ∈ A j Every basic feasible solution corresponds to a policy π . Oliver Friedmann (LMU) Zadeh Lower Bound 11

  11. Games and Policy Iteration Variables of the primal LP t 6 x a is the expected number of x 2 = 1 1 times action a is used, summed 3 2 3 over all starting states. x 4 = 2 2 0 x 1 = 0 x 3 = 0 -4 0 x 5 = 0 x 6 = 1 -1 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 12

  12. Games and Policy Iteration Variables of the primal LP t 6 x a is the expected number of x 2 = 1 1 times action a is used, summed 3 2 3 over all starting states. x 4 = 2 2 0 We have: x 1 = 0 � � r a x π x 3 = 0 val π ( i ) = a -4 0 a ∈ π i ∈ S x 5 = 0 x 6 = 1 -1 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 12

  13. Games and Policy Iteration From MDP to LP max − 1 + 2 x 1 − 2 x 3 − x 5 x 2 = 1 − 1 3 x 1 + 2 3 x 3 + 2 s.t. 3 x 5 x 4 = 2 − x 3 − x 5 6 t x 6 = 1 − x 5 x 1 , x 2 , x 3 , x 4 , x 5 , x 6 ≥ 0 x 2 = 1 1 3 2 3 2 0 x 4 = 2 x 1 = 0 x 3 = 0 0 -4 x 5 = 0 x 5 x 1 x 6 = 1 -1 -1 { x 2 , x 4 , x 6 } x 3 Oliver Friedmann (LMU) Zadeh Lower Bound 13

  14. Games and Policy Iteration From MDP to LP max 5 − 6 x 2 + 2 x 3 + 3 x 5 s.t. x 1 = 3 − 3 x 2 + 2 x 3 + 2 x 5 x 4 = 2 − x 3 − x 5 6 t x 6 = 1 − x 5 x 1 , x 2 , x 3 , x 4 , x 5 , x 6 ≥ 0 x 2 = 0 1 3 2 3 6 6 x 4 = 2 x 1 = 3 x 5 x 3 = 0 0 -4 x 5 = 0 { x 1 , x 4 , x 6 } x 3 x 2 x 6 = 1 -1 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 13

  15. Games and Policy Iteration From MDP to LP max 9 − 6 x 2 − 2 x 4 + x 5 s.t. x 1 = 7 − 3 x 2 − 2 x 4 x 3 = 2 − x 4 − x 5 6 t x 6 = 1 − x 5 x 1 , x 2 , x 3 , x 4 , x 5 , x 6 ≥ 0 x 2 = 0 1 3 2 3 6 6 x 4 = 0 x 1 = 7 x 5 x 3 = 2 2 -4 { x 1 , x 3 , x 6 } x 5 = 0 x 4 x 6 = 1 1 -1 x 2 Oliver Friedmann (LMU) Zadeh Lower Bound 13

  16. Games and Policy Iteration From MDP to LP max 10 − 6 x 2 − 2 x 4 − x 6 s.t. x 1 = 7 − 3 x 2 − 2 x 4 x 3 = 1 − x 4 + x 6 6 t x 5 = 1 − x 6 x 1 , x 2 , x 3 , x 4 , x 5 , x 6 ≥ 0 x 2 = 0 1 3 2 3 6 6 x 4 = 0 { x 1 , x 3 , x 5 } x 4 x 1 = 7 x 2 x 6 x 3 = 1 2 -4 x 5 = 1 x 6 = 0 2 -1 Oliver Friedmann (LMU) Zadeh Lower Bound 13

  17. Games and Policy Iteration Diameter Question: theoretically possible to have polynomially many iterations? Let G be a Markov decision process and n be the number of nodes. Definition: the diameter of G is the least number of iterations required to solve G Small Diameter Theorem The diameter of G is less or equal to n . Oliver Friedmann (LMU) Zadeh Lower Bound 14

  18. Lower Bound for Zadeh’s Rule Lower Bound for Zadeh’s Rule Oliver Friedmann (LMU) Zadeh Lower Bound 15

  19. Lower Bound for Zadeh’s Rule Lower bound construction We define a family of lower bound MDPs G n such that the Least-Entered pivoting rule will simulate an n -bit binary counter. Oliver Friedmann (LMU) Zadeh Lower Bound 16

  20. Lower Bound for Zadeh’s Rule Lower bound construction We define a family of lower bound MDPs G n such that the Least-Entered pivoting rule will simulate an n -bit binary counter. We make use of exponentially growing rewards (and penalties): To get a higher reward the MDP is willing to sacrifice everything that has been built up so far. Oliver Friedmann (LMU) Zadeh Lower Bound 16

  21. Lower Bound for Zadeh’s Rule Lower bound construction We define a family of lower bound MDPs G n such that the Least-Entered pivoting rule will simulate an n -bit binary counter. We make use of exponentially growing rewards (and penalties): To get a higher reward the MDP is willing to sacrifice everything that has been built up so far. Notation: Integer priority p corresponds to reward ( − N ) p , where N = 7 n + 1. . . . < 5 < 3 < 1 < 2 < 4 < 6 < . . . ( − N ) 5 5 for Oliver Friedmann (LMU) Zadeh Lower Bound 16

  22. Lower Bound for Zadeh’s Rule Background The use of priorities is inspired by parity games . Oliver Friedmann (LMU) Zadeh Lower Bound 17

  23. Lower Bound for Zadeh’s Rule Background The use of priorities is inspired by parity games . Friedmann (2009): The strategy iteration algorithm may require exponentially many iterations to solve parity games. Fearnley (2010): The strategy iteration algorithm may require exponentially many iterations to solve MDPs. Oliver Friedmann (LMU) Zadeh Lower Bound 17

  24. Lower Bound for Zadeh’s Rule Background The use of priorities is inspired by parity games . Friedmann (2009): The strategy iteration algorithm may require exponentially many iterations to solve parity games. Fearnley (2010): The strategy iteration algorithm may require exponentially many iterations to solve MDPs. We also first proved a lower bound for parity games and then transferred the result to MDPs and linear programs. Oliver Friedmann (LMU) Zadeh Lower Bound 17

  25. Lower Bound for Zadeh’s Rule Related game-theoretic settings LP-type problems Abstract Concrete Turn-based stochastic games Linear programming 2 1 / 2 players Mean payoff games Markov decision problems 2 players 1 1 / 2 players Parity games Deterministic MDPs 2 players 1 player Oliver Friedmann (LMU) Zadeh Lower Bound 18

  26. Lower Bound for Zadeh’s Rule Related game-theoretic settings LP-type problems Abstract Concrete Turn-based stochastic games Linear programming 2 1 / 2 players Mean payoff games Markov decision problems 2 players 1 1 / 2 players ∈ NP ∩ co NP ∈ P Parity games Deterministic MDPs 2 players 1 player Oliver Friedmann (LMU) Zadeh Lower Bound 18

  27. Lower Bound for Zadeh’s Rule Zadeh’s pivoting rule Zadeh’s Least-Entered rule Perform single switch that has been applied least often. (taken from David Avis’ paper) Oliver Friedmann (LMU) Zadeh Lower Bound 19

  28. Lower Bound for Zadeh’s Rule Tie-Breaking Rule Tie-Breaking Rule = method of selecting a switch in case of a tie (w.r.t. the occurrence record) Proof of Small Diameter Theorem implies: Corollary There is a tie-breaking rule s.t. Zadeh’s rule requires linearly many iterations in the worst-case. Consequence: lower bound construction is equipped with particular tie-breaking rule Oliver Friedmann (LMU) Zadeh Lower Bound 20

  29. Lower Bound for Zadeh’s Rule Binary Counting 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  30. Lower Bound for Zadeh’s Rule Binary Counting 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 TPrinciple: If a bit can be set, then all bits can be set. y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  31. Lower Bound for Zadeh’s Rule Binary Counting 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 1 T Tie-Breaking: We decide to set the first bit. y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  32. Lower Bound for Zadeh’s Rule Binary Counting 0 0 1 R 0 0 1 1 0 0 1 1 0 0 1 1 T Set the second bit and reset the first bit. y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  33. Lower Bound for Zadeh’s Rule Binary Counting 0 0 1 0 0 0 1 0 0 0 1 1 0 0 1 0 T Set the first bit again. y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  34. Lower Bound for Zadeh’s Rule Binary Counting 0 0 1 1 0 0 1 1 0 0 1 2 0 0 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  35. Lower Bound for Zadeh’s Rule Binary Counting 0 1 R R 0 1 1 1 0 1 1 2 0 1 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  36. Lower Bound for Zadeh’s Rule Binary Counting 0 1 0 0 0 1 0 0 0 1 1 2 0 1 0 0 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  37. Lower Bound for Zadeh’s Rule Binary Counting 0 1 0 1 0 1 0 1 0 1 1 3 0 1 0 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  38. Lower Bound for Zadeh’s Rule Binary Counting 0 1 1 R 0 1 1 1 0 1 2 3 0 1 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  39. Lower Bound for Zadeh’s Rule Binary Counting 0 1 1 0 0 1 1 0 0 1 2 3 0 1 1 0 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  40. Lower Bound for Zadeh’s Rule Binary Counting 0 1 1 1 0 1 1 1 0 1 2 4 0 1 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  41. Lower Bound for Zadeh’s Rule Binary Counting 1 R R R 1 1 1 1 1 1 2 4 1 1 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  42. Lower Bound for Zadeh’s Rule Binary Counting 1 0 0 0 1 0 0 0 1 1 2 4 1 0 0 0 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  43. Lower Bound for Zadeh’s Rule Binary Counting 1 0 0 1 1 0 0 1 1 1 2 5 1 0 0 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  44. Lower Bound for Zadeh’s Rule Binary Counting 1 0 1 R 1 0 1 1 1 1 3 5 1 0 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  45. Lower Bound for Zadeh’s Rule Binary Counting 1 0 1 0 1 0 1 0 1 1 3 5 1 0 1 0 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  46. Lower Bound for Zadeh’s Rule Binary Counting 1 0 1 1 1 0 1 1 1 1 3 6 1 0 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  47. Lower Bound for Zadeh’s Rule Binary Counting 1 1 R R 1 1 1 1 1 2 3 6 1 1 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  48. Lower Bound for Zadeh’s Rule Binary Counting 1 1 0 0 1 1 0 0 1 2 3 6 1 1 0 0 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  49. Lower Bound for Zadeh’s Rule Binary Counting 1 1 0 1 1 1 0 1 1 2 3 7 1 1 0 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  50. Lower Bound for Zadeh’s Rule Binary Counting 1 1 1 R 1 1 1 1 1 2 4 7 1 1 1 1 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  51. Lower Bound for Zadeh’s Rule Binary Counting 1 1 1 0 1 1 1 0 1 2 4 7 1 1 1 0 T Continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  52. Lower Bound for Zadeh’s Rule Binary Counting 1 1 1 1 1 1 1 1 1 2 4 8 1 1 1 1 T Problem: Occurrence record unbalanced! y Oliver Friedmann (LMU) Zadeh Lower Bound 21

  53. Lower Bound for Zadeh’s Rule Binary Counting (... again!) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 TLet’s do it again - watch the occurrence record this time! y Oliver Friedmann (LMU) Zadeh Lower Bound 22

  54. Lower Bound for Zadeh’s Rule Binary Counting (... again!) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T Everything okay so far... y Oliver Friedmann (LMU) Zadeh Lower Bound 22

  55. Lower Bound for Zadeh’s Rule Binary Counting (... again!) 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 1 T Everything okay so far... y Oliver Friedmann (LMU) Zadeh Lower Bound 22

  56. Lower Bound for Zadeh’s Rule Binary Counting (... again!) 0 0 1 R 0 0 1 1 0 0 1 1 0 0 1 1 T Everything okay so far... y Oliver Friedmann (LMU) Zadeh Lower Bound 22

  57. Lower Bound for Zadeh’s Rule Binary Counting (... again!) 0 0 1 0 0 0 1 0 0 0 1 1 0 0 1 0 T Problem: We have to set one of the higher bits now! y Oliver Friedmann (LMU) Zadeh Lower Bound 22

  58. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 TReplace gadget by two-bit, conjunctive structure. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  59. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 T Gadget is set iff both edges are going in. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  60. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 T Set one improving edge of every gadget. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  61. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0 1 0 0 0 1 0 0 0 1 T Set other improving edge of first gadget. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  62. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 0 1 0 0 0 1 1 1 1 1 0 0 0 1 0 0 0 1 0 0 0 1 T Other gadgets have updated to their old setting. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  63. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 0 1 0 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 T Set one improving edge of every gadget again. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  64. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 1 R 0 0 1 1 1 1 2 1 0 0 1 1 1 1 1 1 1 1 1 1 T Set other improving edge of second gadget. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  65. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 1 0 0 0 1 0 1 1 2 1 0 0 1 0 1 1 1 1 0 0 1 0 T Reset all other gadgets. y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  66. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 1 0 1 1 1 1 2 2 2 2 0 0 1 0 1 1 1 1 0 0 1 0 T Everything okay so far, continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  67. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 1 1 1 1 1 1 2 2 2 2 0 0 1 1 1 1 1 2 0 0 1 1 T Everything okay so far, continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 23

  68. Lower Bound for Zadeh’s Rule Binary Counting with conjunctive bits 0 0 1 1 0 0 1 1 2 2 2 2 0 0 1 1 1 1 1 2 0 0 1 1 T Everything okay so far, continue... y Oliver Friedmann (LMU) Zadeh Lower Bound 23

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Subexponential lower bounds for randomized pivoting rules for the simplex algorithm Oliver

Subexponential lower bounds for randomized pivoting rules for the simplex algorithm Oliver

Subexponential lower bounds for randomized pivoting rules for the simplex algorithm Oliver Friedmann 1 Thomas Dueholm Hansen 2 Uri Zwick 3 1 Department of Computer Science, University of Munich, Germany. 2 Department of Management Science and

957 views • 94 slides
A Superpolynomial Lower Bound for Clique Function Circuits with at most 1 6 loglog n Negation

A Superpolynomial Lower Bound for Clique Function Circuits with at most 1 6 loglog n Negation

Main Results Hawks Eye View Extending Montone Lower Bounds to Limited Negations Monotone Lower bound for Clique function using Bottleneck counting Connecting the two worlds Thank You A Superpolynomial Lower Bound for Clique Function

1.03k views • 71 slides
Sorting Lower Bound Sorting Lower Bound 1 Comparison-Based Sorting (10.4) Many sorting

Sorting Lower Bound Sorting Lower Bound 1 Comparison-Based Sorting (10.4) Many sorting

Sorting Lower Bound Sorting Lower Bound 1 Comparison-Based Sorting (10.4) Many sorting algorithms are comparison based. They sort by making comparisons between pairs of objects Examples: bubble-sort, selection-sort, insertion-sort,

75 views • 5 slides
Upper and Lower Loop Bound Estimation by Symbolic Execution and Loop Acceleration Pavel Cadek

Upper and Lower Loop Bound Estimation by Symbolic Execution and Loop Acceleration Pavel Cadek

Upper and Lower Loop Bound Estimation by Symbolic Execution and Loop Acceleration Pavel Cadek 1 1 TU Wien, Austria Formal Methods in Computer-Aided Design 30 Oct - 2 Nov, 2018 Loop Bound Analysis Upper loop bound: max { n , 0 } Lower loop

248 views • 5 slides
Techniques to lower bound extension complexity Thomas Rothvoss UW Seattle Known lower bounds on

Techniques to lower bound extension complexity Thomas Rothvoss UW Seattle Known lower bounds on

Techniques to lower bound extension complexity Thomas Rothvoss UW Seattle Known lower bounds on extended formulation Kaibel, Razborovs Inform. SA Weltge symmetry arg. theory + Fourier COR/ yes yes yes ? TSP [KW13]

1.25k views • 108 slides
MA Macroeconomics 6. The Zero Lower Bound and the Liquidity Trap Karl Whelan School of

MA Macroeconomics 6. The Zero Lower Bound and the Liquidity Trap Karl Whelan School of

MA Macroeconomics 6. The Zero Lower Bound and the Liquidity Trap Karl Whelan School of Economics, UCD Autumn 2014 Karl Whelan (UCD) The Zero Lower Bound Autumn 2014 1 / 25 Can Interest Rates Be Negative? Up to now, we have assumed that the

493 views • 25 slides
Advanced Macroeconomics 4. The Zero Lower Bound and the Liquidity Trap Karl Whelan School of

Advanced Macroeconomics 4. The Zero Lower Bound and the Liquidity Trap Karl Whelan School of

Advanced Macroeconomics 4. The Zero Lower Bound and the Liquidity Trap Karl Whelan School of Economics, UCD Spring 2020 Karl Whelan (UCD) The Zero Lower Bound Spring 2020 1 / 21 Can Interest Rates Be Negative? Up to now, we have assumed

541 views • 21 slides
Recent progress on a sharp lower bound for first (nonzero) Steklov eigenvalue Chao Xia (Xiamen

Recent progress on a sharp lower bound for first (nonzero) Steklov eigenvalue Chao Xia (Xiamen

Recent progress on a sharp lower bound for first (nonzero) Steklov eigenvalue Chao Xia (Xiamen University) (joint with Changwei Xiong) Asia-Pacific Analysis and PDE Seminar May 11th, 2020 1 / 30 Table of Contents Review of eigenvalue lower

779 views • 32 slides
403: Algorithms and Data Structures Lower Bound for Sor&lt;ng &amp; The closest pair in 2D

403: Algorithms and Data Structures Lower Bound for Sor&lt;ng &amp; The closest pair in 2D

403: Algorithms and Data Structures Lower Bound for Sor&lt;ng &amp; The closest pair in 2D Fall 2016 UAlbany Computer Science So far: Sor&lt;ng Algorithm Time Space Inser&amp;on O(n 2 ) in-place Merge O(n logn) 2 nd array to

407 views • 16 slides
How Quantum Computing This Lower Bound Is . . . Can Help With (Continuous) How Quantum . . .

How Quantum Computing This Lower Bound Is . . . Can Help With (Continuous) How Quantum . . .

Known Advantages of . . . Need to Consider . . . We Consider Only . . . Non-Quantum Lower . . . How Quantum Computing This Lower Bound Is . . . Can Help With (Continuous) How Quantum . . . Auxiliary Quantum . . . Optimization Main

285 views • 24 slides
Lecture 8: SOS Lower Bound for 3-XOR Lecture Outline Part I: SOS Lower Bounds from Pseudo-

Lecture 8: SOS Lower Bound for 3-XOR Lecture Outline Part I: SOS Lower Bounds from Pseudo-

Lecture 8: SOS Lower Bound for 3-XOR Lecture Outline Part I: SOS Lower Bounds from Pseudo- expectation Values Part II: Random 3-XOR Equations and Pseudo- expectation Values Part III: Proving PSDness Part IV: Analyzing Parameter

650 views • 37 slides
Permanent V. Determinant: An Exponential Lower Bound Assuming Symmetry J.M. Landsberg and

Permanent V. Determinant: An Exponential Lower Bound Assuming Symmetry J.M. Landsberg and

Permanent V. Determinant: An Exponential Lower Bound Assuming Symmetry J.M. Landsberg and Nicolas Ressayre Texas A&amp;M University and Univ. Lyon I ITCS 2016 1 / 8 Valiants conjecture Theorem (Valiant) Let P be a homogeneous polynomial

220 views • 8 slides
Sorting Lower Bound Radix Sort Radix sort to the rescue sort of

Sorting Lower Bound Radix Sort Radix sort to the rescue sort of

What is the min height of a tree with X external nodes? Sorting Lower Bound Radix Sort Radix sort to the rescue sort of http://www.cs.auckland.ac.nz/software/AlgAnim/radixsort.html EditorTree evals due last night late is better

554 views • 14 slides
Linear lower bound on degrees of Positivstellensatz calculus proofs for the parity (Dima

Linear lower bound on degrees of Positivstellensatz calculus proofs for the parity (Dima

Linear lower bound on degrees of Positivstellensatz calculus proofs for the parity (Dima Grigoriev) talk by Anastasia Sofronova Seminar Modern Methods in CS April 21, 2020 1 / 17 Proof systems an unsatisfiable formula in CNF.

597 views • 23 slides
Solving Ax=b with Pivoting Solving Ax=b with Gaussian Elimination and LU and partial

Solving Ax=b with Pivoting Solving Ax=b with Gaussian Elimination and LU and partial

Numerical and Scientific Computing with Applications David F . Gleich CS 314, Purdue September 29, 2016 In this class: Solving Ax=b with Pivoting Solving Ax=b with Gaussian Elimination and LU and partial pivoting Next class

240 views • 3 slides
An asymptotic lower bound for the norm of the Laplace operator on a space of polynomials

An asymptotic lower bound for the norm of the Laplace operator on a space of polynomials

An asymptotic lower bound for the norm of the Laplace operator on a space of polynomials Christian Rebs Technische Universit at Chemnitz, 17 . 8 . 2017 Introduction A. B ottcher, C.Rebs: On the constants in Markov inequalities for the

710 views • 17 slides
Uni.lu HPC School 2019 PS14: Distributed Mixed-Integer Programming (MIP) optimization with Cplex

Uni.lu HPC School 2019 PS14: Distributed Mixed-Integer Programming (MIP) optimization with Cplex

Uni.lu HPC School 2019 PS14: Distributed Mixed-Integer Programming (MIP) optimization with Cplex and Gurobi Uni.lu High Performance Computing (HPC) Team E. Kieffer University of Luxembourg (UL), Luxembourg http://hpc.uni.lu E. Kieffer &amp;

374 views • 9 slides
P a r t 1 7 L i n e a r p r o g r a m m i n g 2 : A n a v e s o

P a r t 1 7 L i n e a r p r o g r a m m i n g 2 : A n a v e s o

P a r t 1 7 L i n e a r p r o g r a m m i n g 2 : A n a v e s o l u t i o n a l g o r i t h m T x minimize c Ax b 6 1 A nave algorithm Theorem: A polyhedron can only have finitely many

588 views • 47 slides
Effective Linear Programming-Based Placement Techniques Sherief Reda Sherief Reda Amit

Effective Linear Programming-Based Placement Techniques Sherief Reda Sherief Reda Amit

Effective Linear Programming-Based Placement Techniques Sherief Reda Sherief Reda Amit Chowdhary Amit Chowdhary UC San Diego UC San Diego Intel Corporation Intel Corporation Outline Outline Motivation Motivation Modeling of

296 views • 27 slides
Module 8 Linear Programming CS 886 Sequential Decision Making and Reinforcement Learning

Module 8 Linear Programming CS 886 Sequential Decision Making and Reinforcement Learning

Module 8 Linear Programming CS 886 Sequential Decision Making and Reinforcement Learning University of Waterloo Policy Optimization Value and policy iteration Iterative algorithms that implicitly solve an optimization problem Can

423 views • 16 slides
Support Vector Machines Support Vector Machines Hypothesis Space Hypothesis Space variable

Support Vector Machines Support Vector Machines Hypothesis Space Hypothesis Space variable

Support Vector Machines Support Vector Machines Hypothesis Space Hypothesis Space variable size variable size deterministic deterministic continuous parameters continuous parameters Learning Algorithm Learning

546 views • 22 slides
r rr t

r rr t

r rr t tts rst s rr

424 views • 8 slides
Introduction to Linear Programming Consider the Diet problem: - n food items, m nutrients - for

Introduction to Linear Programming Consider the Diet problem: - n food items, m nutrients - for

Introduction to Linear Programming Consider the Diet problem: - n food items, m nutrients - for every nutrient: the daily quota b j - for each item: cost per pound c i - for every item and nutrient: how much of the nutrient in a pound of item: a

401 views • 6 slides
Bilevel Integer Programming Ted Ralphs 1 Joint work with: Scott DeNegre 1 , Menal Guzelsoy 2 ,

Bilevel Integer Programming Ted Ralphs 1 Joint work with: Scott DeNegre 1 , Menal Guzelsoy 2 ,

Bilevel Integer Programming Ted Ralphs 1 Joint work with: Scott DeNegre 1 , Menal Guzelsoy 2 , Andrea Lodi 3 , Fabrizio Rossi 4 , Stefano Smriglio 4 1 COR@L Lab, Department of Industrial and Systems Engineering, Lehigh University 2 ISyE, Georgia

564 views • 40 slides

More recommend