Presentation is loading. Please wait.

Presentation is loading. Please wait.

Discrete Optimization Last Time  Parallel Machine Scheduling  Lagrange Relaxation for Flow Shop Scheduling Today  Local Search  Meta-Heuristic Search.

Similar presentations


Presentation on theme: "Discrete Optimization Last Time  Parallel Machine Scheduling  Lagrange Relaxation for Flow Shop Scheduling Today  Local Search  Meta-Heuristic Search."— Presentation transcript:

1 Discrete Optimization Last Time  Parallel Machine Scheduling  Lagrange Relaxation for Flow Shop Scheduling Today  Local Search  Meta-Heuristic Search  Specific Search Strategies  Simulated Annealing  Stochastic Machines  Convergence Analysis 2015/4/25 Shi-Chung Chang, NTUEE, GIIE, GICE, Spring, 2008

2 Reading Assignments S. Kirkpatrick and C. D. Gelatt and M. P. Vecchi, Optimization by Simulated Annealing, Science, Vol 220, Number 4598, pages , "General Simulated Annealing Algorithm""General Simulated Annealing Algorithm" An open-source MATLAB program for general simulated annealing exercises File.do?objectId=10548&objectType=file 2015/4/25Shi-Chung Chang, NTUEE, GIIE, GICE

3 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.3 Copyright © 2007 by Leong Hon Wai Combinatorial Optimization Combinatorial Optimization Problem:  Consists of (R, C), where uR is a set of configuration uC : R  , is a cost function  Given (R, C), find s*  R, such that C(s*) = min s  R { C(s) } Example 1: Travelling Salesman Problem (TSP) Given n cities, and distance matrix [d ij ] To find: shortest tour of n cities (visit each city exactly once) R = { all cyclic permutations  of the n cities }

4 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.4 Copyright © 2007 by Leong Hon Wai Combinatorial Optimization Example 3: Linear Programming (LP) Example 2: Minimum Spanning Tree Problem (MST) Given: G = (V, E), and symmetric distance matrix [d ij ] To find: spanning tree T of G with minimum total edge cost R = { T : T =(V, E’) is a spanning tree of G }

5 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.5 Copyright © 2007 by Leong Hon Wai Some Common Types of COP  TSP, Quadratic Assignment Problem  Bin Packing and Generalised Assignment Problems  Hub allocation problems  Graph Colouring & Partitioning  Vehicle Routing  Single & Multiple Knapsack  Set Partitioning & Set Covering Problems  Processor Allocation Problem  Various Staff Scheduling Problems  Job Shop Scheduling

6 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.6 Copyright © 2007 by Leong Hon Wai Outline Today  Local Search  Meta-Heuristic Search  Specific Search Strategies  Simulated Annealing  Stochastic Machines  Convergence Analysis 2015/4/25 Shi-Chung Chang, NTUEE, GIIE, GICE, Spring, 2008

7 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.7 Copyright © 2007 by Leong Hon Wai Techniques for Tackling COPs  COPs often formulated as Integer Linear Programs (ILPs)  But far too slow to solve them this way  Operations Research  Branch & Bound, Cutting Plane Algorithms  A.I. (excluding neural nets)  A*, Constraint Logic Programming  All these techniques can guarantee an optimal solution, but suffer from exponential runtime in the worst case

8 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.8 Copyright © 2007 by Leong Hon Wai Techniques for Tackling COPs  Heuristics solutions (do not guarantee optimal)  Tailored heuristics  e.g. Lin Kernigan for TSP, GP  In general: run very fast, find good solutions  But cannot be applied to other problems  Meta-heuristic search algorithms (MHs)  Guide the local search for new (improved) solutions  Hill climbing is a kind of local search, u may form part of a metaheuristic

9 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.9 Copyright © 2007 by Leong Hon Wai Neighborhood Search Algorithms  Start with a feasible solution x  Define a neighborhood of x  Identify an improved neighbor y  Replace x by y and repeat x x x x x x x x x x x x x x xx xx x x x x x xx x x x x x x x x xx x x x x x x x x x x x x x x x x x xx xx x x x x x xx x x x x x x x x xx x x x x x x x x x x x x x x x x x xx xx x x x x x xx x x x x x x x x xx x x x x x x x x x x x x x x x x x xx xx x x x x x xx x x x x x x x x xx x x x ……………..

10 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.10 Copyright © 2007 by Leong Hon Wai Generic Neighbourhood Search Generic Neighbourhood_Search; begin s  initial solution s0; repeat choose s’  N(s); s  s’; until (Terminating condition); end;

11 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.11 Copyright © 2007 by Leong Hon Wai Example: Map Coloring (1) 1. Start with random coloring of nodes 2. Change color of one node to reduce # of con 3. Repeat 2

12 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.12 Copyright © 2007 by Leong Hon Wai Example: Graph Colouring (2) #conflicts unchanged, but different solution.

13 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.13 Copyright © 2007 by Leong Hon Wai Neighbourhood Search (history…)  Neighbourhood Search dates back a long way  Simplex Algorithm for LP Viewed as a search technique  KL / FM algorithm for Graph Partitioning  Augmenting path algorithms  maximum flow in networks  bipartite matching  non-bipartite matching  Iteratively improves the current solution.

14 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.14 Copyright © 2007 by Leong Hon Wai Shape of the Cost Function (1-D)

15 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.15 Copyright © 2007 by Leong Hon Wai Shape of the Cost Function (2-D)  Multiple Local Optima  Plateaus  Ridges: going up only in a narrow direction.

16 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.16 Copyright © 2007 by Leong Hon Wai Outline Today  Local Search  Meta-Heuristic Search  Specific Search Strategies  Simulated Annealing  Stochastic Machines  Convergence Analysis 2015/4/25 Shi-Chung Chang, NTUEE, GIIE, GICE, Spring, 2008

17 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.17 Copyright © 2007 by Leong Hon Wai Meta-Heuristic Search… Design Issues  Solution Representation  Initial Solution (where to start)  Neighbourhood (strength & size)  Cost Function (viz-a-viz objective function)  Search strategy (how to move)  Refinements  Termination (when to stop)

18 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.18 Copyright © 2007 by Leong Hon Wai Local Search  A local search procedure looks for the “best solution” that is “near” another solution  by repeatedly making local changes to current soln  until no further improved solutions can be found  Local search procedures are generally problem specific  Most MHs use some kind of local search procedure to actually perform the search

19 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.19 Copyright © 2007 by Leong Hon Wai Don’t Settle for Second Best  MHs are generally attracted to good solutions – local optima  Simple MHs find it difficult or impossible to escape these attractive points in the search space  All of the popular MHs employed today have some mechanism(s) for escaping local optima  These will be discussed for each MH we cover today

20 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.20 Copyright © 2007 by Leong Hon Wai Ways of Getting Around (in Search Space)  MHs can be roughly categorized into  Iterative  Population-based  Constructive  Of course, in an attempt to prove by exhaustive search that there really are an infinite number of research papers…  Researches have combined these broad approaches in many and varied ways

21 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.21 Copyright © 2007 by Leong Hon Wai Popular Iterative MHs  Iterative MHs start with one or more feasible solutions  Apply transitions to reach new solutions (i.e. local search)  Transitions include move, swap, add, drop, invert plus others  The neighbourhood ( N(X) ) of a solution consists of those solutions reachable by applying (generally) one transition  Simulated Annealing  Tabu Search

22 Hon Wai Leong, NUS (CS5234, 13 Nov 2007) Page L13.22 Copyright © 2007 by Leong Hon Wai Outline Today  Local Search  Meta-Heuristic Search  Specific Search Strategies  Simulated Annealing  Stochastic Machines  Convergence Analysis 2015/4/25 Shi-Chung Chang, NTUEE, GIIE, GICE, Spring, 2008

23 CS 561, Session 7 23 Search Strategies Uninformed: Use only information available in the problem formulation Breadth-first Uniform-cost Depth-first Depth-limited Iterative deepening Informed: Use heuristics to guide the search Best first A*

24 CS 561, Session 7 24 Evaluation of search strategies Search algorithms are commonly evaluated according to the following four criteria: Completeness: does it always find a solution if one exists? Time complexity: how long does it take as a function of number of nodes? Space complexity: how much memory does it require? Optimality: does it guarantee the least-cost solution? Time and space complexity are measured in terms of: b – max branching factor of the search tree d – depth of the least-cost solution m – max depth of the search tree (may be infinity)

25 CS 561, Session 7 25 Informed search Use heuristics to guide the search Best first A* Heuristics Hill-climbing Simulated annealing

26 CS 561, Session 7 26 Best-first search Idea: use an evaluation function for each node; estimate of “desirability”  expand most desirable unexpanded node. Implementation: QueueingFn = insert successors in decreasing order of desirability Special cases: greedy search A* search

27 CS 561, Session 7 27 Romania with step costs in km

28 CS 561, Session 7 28 Greedy search Estimation function: h(n) = estimate of cost from n to goal (heuristic) For example: h SLD (n) = straight-line distance from n to Bucharest Greedy search expands first the node that appears to be closest to the goal, according to h(n).

29 CS 561, Session 7 29

30 CS 561, Session 7 30

31 CS 561, Session 7 31

32 CS 561, Session 7 32

33 CS 561, Session 7 33 Properties of Greedy Search Complete? Does it always give a solution if one exists? Time? How long does it take? Space? How much memory is needed? Optimal? Does it give the optimal path?

34 CS 561, Session 7 34 A* search Idea: combine greedy approach and avoid expanding paths that are already expensive evaluation function: f(n) = g(n) + h(n)with: g(n) – cost so far to reach n h(n) – estimated cost to goal from n f(n) – estimated total cost of path through n to goal A* search uses an admissible heuristic, that is, h(n)  h*(n) where h*(n) is the true cost from n. For example: h SLD (n) never overestimates actual road distance. Theorem: A* search is optimal

35 CS 561, Session 7 35

36 CS 561, Session 7 36

37 CS 561, Session 7 37

38 CS 561, Session 7 38

39 CS 561, Session 7 39

40 CS 561, Session 7 40

41 CS 561, Session 7 41 Optimality of A* (standard proof) Suppose some suboptimal goal G 2 has been generated and is in the queue. Let n be an unexpanded node on a shortest path to an optimal goal G 1. (since g(G1)>f(n))

42 CS 561, Session 7 42 Optimality of A* (more useful proof)

43 CS 561, Session 7 43 f-contours How do the contours look like when h(n) =0?

44 CS 561, Session 7 44 Properties of A* Complete? Time? Space? Optimal?

45 CS 561, Session 7 45 Proof of lemma: pathmax

46 CS 561, Session 7 46 Admissible heuristics

47 CS 561, Session 7 47 Relaxed Problem How to determine an admissible heuristics? E.g. h1 and h2 in the 8-puzzle problem Admissible heuristics can be derived from the exact solution cost of a relaxed version of the problem. Example: A tile can move from square A to square B If A is adjacent to B and If B in blank Possible relaxed problems A tile can move from square A to square B if A is adjacent to B A tile can move from square A to square B if B is blank A tile can move from square A to square B

48 CS 561, Session 7 48 Next Iterative improvement Hill climbing Simulated annealing

49 CS 561, Session 7 49 Iterative improvement In many optimization problems, path is irrelevant; the goal state itself is the solution. Then, state space = space of “complete” configurations. Algorithm goal: - find optimal configuration (e.g., TSP), or, - find configuration satisfying constraints (e.g., n- queens) In such cases, can use iterative improvement algorithms: keep a single “current” state, and try to improve it.

50 CS 561, Session 7 50 Iterative improvement example: n-queens Goal: Put n chess-game queens on an n x n board, with no two queens on the same row, column, or diagonal. Here, goal state is initially unknown but is specified by constraints that it must satisfy.

51 CS 561, Session 7 51 Hill climbing (or gradient ascent/descent) Iteratively maximize “value” of current state, by replacing it by successor state that has highest value, as long as possible.

52 CS 561, Session 7 52 Hill climbing Note: minimizing a “value” function v(n) is equivalent to maximizing –v(n), thus both notions are used interchangeably. Notion of “extremization”: find extrema (minima or maxima) of a value function.

53 CS 561, Session 7 53 Hill climbing Problem: depending on initial state, may get stuck in local extremum. Any suggestion? Does it matter if you start from left or from right?

54 Outline Today Local Search Meta-Heuristic Search Specific Search Strategies Simulated Annealing Stochastic Machines Convergence Analysis 2015/4/25 Shi-Chung Chang, NTUEE, GIIE, GICE, Spring, 2008

55 CS 561, Session 7 55 Minimizing energy new formalism: - let’s compare our state space to that of a physical system that is subject to natural interactions, - and let’s compare our value function to the overall potential energy E of the system. On every updating we have  E  0

56 CS 561, Session 7 56 Minimizing energy On every updating we have  E  0 Hence the dynamics of the system tend to move E toward a minimum. Note: There may be different such local minma. Global minimization is not guaranteed.

57 CS 561, Session 7 57 Local Minima Problem Question: How do you avoid this local minima? starting point descend direction local minima global minima barrier to local search

58 CS 561, Session 7 58 Consequences of the Occasional Ascents Help escaping the local optima. desired effect Might pass global optima after reaching it adverse effect

59 CS 561, Session 7 59 Boltzmann machines h The Boltzmann Machine of Hinton, Sejnowski, and Ackley (1984) uses simulated annealing to escape local minima. consider how one might get a ball-bearing traveling along the curve to "probably end up" in the deepest minimum. The idea is to shake the box "about h hard" — then the ball is more likely to go from D to C than from C to D. So, on average, the ball should end up in C's valley.

60 CS 561, Session 7 60 Question: What is the difference between this problem and our problem (finding global minima)?

61 CS 561, Session 7 61 Simulated annealing: basic idea From current state, pick a random successor state: If it has better value than current state, then “accept the transition,” that is, use successor state as current state; Otherwise, instead flip a coin and accept the transition with a given probability (that is lower as the successor is worse). So we accept to sometimes “un-optimize” the value function a little with a non-zero probability.

62 CS 561, Session 7 62 Boltzmann’s statistical theory of gases In the statistical theory of gases, the gas is described not by a deterministic dynamics, but rather by the probability that it will be in different states. The 19th century physicist Ludwig Boltzmann developed a theory that included a probability distribution of temperature (i.e., every small region of the gas had the same kinetic energy).

63 CS 561, Session 7 63 Boltzmann distribution At thermal equilibrium at temperature T, the Boltzmann distribution gives the relative probability that the system will occupy state A vs. state B as: where E(A) and E(B) are the energies associated with states A and B.

64 CS 561, Session 7 64 Simulated annealing Kirkpatrick et al. 1983: Simulated annealing is a general method for making likely the escape from local minima by allowing jumps to higher energy states. The analogy here is with the process of annealing used by a craftsman in forging a sword from an alloy.

65 CS 561, Session 7 65 Simulated annealing: Sword He heats the metal, then slowly cools it as he hammers the blade into shape. If he cools the blade too quickly the metal will form patches of different composition; If the metal is cooled slowly while it is shaped, the constituent metals will form a uniform alloy.

66 CS 561, Session 7 66 Simulated annealing: Sword He heats the metal, then slowly cools it as he hammers the blade into shape. If he cools the blade too quickly the metal will form patches of different composition; If the metal is cooled slowly while it is shaped, the constituent metals will form a uniform alloy. Example: arranging cubes in a box (sugure cubes)

67 CS 561, Session 7 67 Simulated annealing in practice -set T -optimize for given T -lower T -Repeat (see Geman & Geman, 1984)

68 CS 561, Session 7 68 Simulated annealing in practice Geman & Geman (1984): if T is lowered sufficiently slowly (with respect to the number of iterations used to optimize at a given T), simulated annealing is guaranteed to find the global minimum. Caveat: this algorithm has no end (Geman & Geman’s T decrease schedule is in the 1/log of the number of iterations, so, T will never reach zero), so it may take an infinite amount of time for it to find the global minimum.

69 CS 561, Session 7 69 Simulated annealing algorithm Idea: Escape local extrema by allowing “bad moves,” but gradually decrease their size and frequency. Algorithm when goal is to minimize E. < - -

70 CS 561, Session 7 70 Note on simulated annealing: limit cases Boltzmann distribution: accept “bad move” with  E<0 (goal is to maximize E) with probability P(  E) = exp(  E/T) If T is large:  E < 0  E/T < 0 and very small exp(  E/T) close to 1 accept bad move with high probability If T is near 0:  E < 0  E/T < 0 and very large exp(  E/T) close to 0 accept bad move with low probability Random walk Deterministic down-hill

71 CS 561, Session 7 71 Summary A* search = best-first with measure = path cost so far + estimated path cost to goal. - combines advantages of uniform-cost and greedy searches - complete, optimal and optimally efficient - space complexity still exponential (Skip)

72 CS 561, Session 7 72 Summary Time complexity of heuristic algorithms depend on quality of heuristic function. Good heuristics can sometimes be constructed by examining the problem definition or by generalizing from experience with the problem class. Iterative improvement algorithms keep only a single state in memory. Can get stuck in local extrema; simulated annealing provides a way to escape local extrema, and is complete and optimal given a slow enough cooling schedule.

73 Outline Today Local Search Meta-Heuristic Search Specific Search Strategies Simulated Annealing Stochastic Machines Convergence Analysis 2015/4/25 Shi-Chung Chang, NTUEE, GIIE, GICE, Spring, 2008

74 Stochastic Machines CS679 Lecture Note by Jin Hyung Kim Computer Science Department KAIST

75 Statistical Machine Root at statistical mechanics  derive thermodynamic properties of macroscopic bodies from microscopic elements probabilistic nature due to enormous degree of freedom  concept of entropy plays the central role  Gibbs distribution Markov Chain Metropolis algorithm Simulated Annealing Boltzman Machine  device for modeling the underlying probability distribution of data set

76 Statistical Mechanics In thermal equilibrium, probability of state i  energy of state i  absolute temperature  Boltzman constant

77 Markov Chain Stochastic process of Markov property state X n+1 at time n+1depends only on state X n Transition probability & Stochastic matrix m-step transition probability

78 Markov Chain Recurrent state  P(ever returning to the state i) = 1 Transient state  P(ever returning to the state i) < 1 mean recurrence time of state i : T i (k)  expectation of time elapsed between (k-1) th return to k th return steady-state probability of state i,  i   I = 1/(mean recurrence time) ergodicity  long-term proportion of time spent in state i approaches to the steady-state probability

79 Convergence to stationary distribution State distribution vector starting from arbitrary initial distribution, transition prob will converge to stationary distribution for ergodic Markov chain  independent of initial distribution

80 Metropolis algorithm Modified Monte Carlo method Suppose our objective is to reach the state minimizing energy function 1. Randomly generate a new state, Y, from state X 2. If  E(energy difference between Y and X) < 0 then move to Y (set Y to X) and goto 1 3. Else  3.1 select a random number,   3.2 if  < exp(-  E / T) then move to Y (set Y to X) and goto 1  3.3 else goto 1

81 Metropolis algrthm and Markov Chain choose probability distribution so that Markov chain converge to be a Gibbs distribution then where Metropolis algorithm is equivalent to random step in stationary Markov chain  shown that such choice satisfied principle of detailed balance

82 Simulated Annealing Solves combinatorial optimization variant of Metropolis algorithm by S. Kirkpatric (83) finding minimum-energy solution of a neural network = finding low temperature state of physical system To overcome local minimum problem Key idea  Instead always going downhill, try to go downhill ‘most of the time’

83 Iterative + Statistical Simple Iterative Algorithm (TSP) 1. find a path p 2. make p’, a variation of p 3. if p’ is better than p, keep p’ as p 4. goto 2 Metropolis Algorithm  3’ : if (p’ is better than p) or (random < Prob), then keep p’ as p  a kind of Monte Carlo method Simulated Annealing  T is reduced as time passes

84 About T Metropolis Algorithm  Prob = p(  E) = exp (  E / T) Simulated Annealing  Prob = p i (  E) = exp (  E / T i )  if T i is reduced too fast, poor quality  if T t >= T(0) / log(1+t) - Geman  System will converge to minimun configuration  T t = k/1+t - Szu  T t =  T(t-1) where  is in between 0.8 and 0.99

85 Function Simulated Annealing current  select a node (initialize) for t  1 to  do T  schedule[t] if T=0 then return current next  a random selected successor of current  E  value[next] - value[current] if  E > 0 then current  next else current  next only with probability e  E /T

86 Outline Today  Local Search  Meta-Heuristic Search  Specific Search Strategies  Simulated Annealing  Stochastic Machines  Convergence Analysis 2015/4/25 Shi-Chung Chang, NTUEE, GIIE, GICE, Spring, 2008

87 Convergence Analysis of Simulated Annealing Ref: E. Aarts, J. Korst, P. Van Laarhoven, “Simulated Annealing,” in Local Search in Combinatorial Optimization, edited by E. Aarts and J. Lenstra, 1997, pp

88 Boltzmann distribution At thermal equilibrium at temperature T, the Boltzmann distribution gives the relative probability that the system will occupy state A vs. state B as: where E(A) and E(B) are the energies associated with states A and B.

89 Simulated annealing in practice Geman & Geman (1984): if T is lowered sufficiently slowly (with respect to the number of iterations used to optimize at a given T), simulated annealing is guaranteed to find the global minimum. Caveat: this algorithm has no end (Geman & Geman’s T decrease schedule is in the 1/log of the number of iterations, so, T will never reach zero), so it may take an infinite amount of time for it to find the global minimum.

90 Simulated annealing algorithm Idea: Escape local extrema by allowing “bad moves,” but gradually decrease their size and frequency. Algorithm when goal is to minimize E. < - -

91 Note on simulated annealing: limit cases Boltzmann distribution: accept “bad move” with  E<0 (goal is to maximize E) with probability P(  E) = exp(  E/T) If T is large:  E < 0  E/T < 0 and very small exp(  E/T) close to 1 accept bad move with high probability If T is near 0:  E < 0  E/T < 0 and very large exp(  E/T) close to 0 accept bad move with low probability Random walk Deterministic down-hill

92 Markov Model Solution  State Cost of a solution  Energy of a state Generation Probability of state j from state i: G ij

93 Acceptance and Transition Probabilities Acceptance probability of state j as next state at state i Transition probability from state i to state j

94 Irreducibility and Aperiodicity of M.C. Irriducibility Aperiodicity

95 Theorem 1: Existence of Unique Stationary State Distribution Finite homogeneous M.C. Irriducibility + Aperiodicity  existence of unique stationary distribution

96 Theorem 2: Asymptotic Convergence of Simulated Annealing P(k): the transition matrix of the homogeneous M.C. associated with the S.A. algorithm C k = C for all k  existence of a unique stationary distribution

97 Asymptotic Convergence of Simulated Annealing From Theorem 2


Download ppt "Discrete Optimization Last Time  Parallel Machine Scheduling  Lagrange Relaxation for Flow Shop Scheduling Today  Local Search  Meta-Heuristic Search."

Similar presentations


Ads by Google