wright university tabu search part1

Upload: zougoulougou

Post on 06-Apr-2018

223 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 Wright University Tabu Search Part1

    1/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Tabu Search: Fundamentals

  • 8/3/2019 Wright University Tabu Search Part1

    2/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Scenario #1:You are playing fetch with your dog

    Challenge:Why does he/she not find the ball if

    you throw it a different direction?

  • 8/3/2019 Wright University Tabu Search Part1

    3/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Scenario #2:You have misplaced your car keys.

    Challenge:How might you go about locating

    them?

  • 8/3/2019 Wright University Tabu Search Part1

    4/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Key Concepts of Scenario

    Avoid self-defeating actions Keep a memory of search experience Thoroughly search promising areas Examine new search areas if necessary Learn from the search

  • 8/3/2019 Wright University Tabu Search Part1

    5/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Background

    According to Fred Glover, these wereconcepts he uncoveredoArose from an AI project course as a studento Trying to get the computer to emulate (human)

    problem solvingo His scenario involved fellow students solving an

    optimization problemo Found they followed a fairly systematic

    approach

  • 8/3/2019 Wright University Tabu Search Part1

    6/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Motivation for Tabu Search

  • 8/3/2019 Wright University Tabu Search Part1

    7/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Tabu Search - An Abstract

    A Meta-HeuristicoCan be used to guide subordinate

    methods Employs adaptive memory functions

    oAvoid repeat solutions using a tabu listo

    Intensifies the search in local regions Employs a responsive exploration functionoDiversifies to new areas of search space

  • 8/3/2019 Wright University Tabu Search Part1

    8/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Why Memory

    Memory-less searcheso Genetic algorithmso Simulated annealingo No explicit use or exploitation of memory

    Rigid memory methodso Branch and bound methodso The solution approach completely dictated

    Inherited memoryo Evolutionary methods (genetic algorithms)

  • 8/3/2019 Wright University Tabu Search Part1

    9/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Resistance to Memory

    TS use of memory not immediatelyaccepted

    Provides many more degrees of freedomo Result is approaches harder to proveo Not as mathematically tractable

    Use of memory also brings pitfallso Proper use of memory is not immediately

    known Large part of the tabu search research nowis finding integrating principles that provideguidelines for best use of memory

  • 8/3/2019 Wright University Tabu Search Part1

    10/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Tabu Search Functions

    Current Focus

  • 8/3/2019 Wright University Tabu Search Part1

    11/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Purpose of Functions

  • 8/3/2019 Wright University Tabu Search Part1

    12/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Short-Term Tabu Search

    Tabu Search moves amongneighborhoods

  • 8/3/2019 Wright University Tabu Search Part1

    13/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    What are Moves?

    Toggle variable between 0 and 1 Swap nodes in a routing tour Swap nodes between tours Insert/delete edge operation on graph Interchange variables from basis

    There are common moves, but generally moves arespecific to the application area. Moves define theneighborhood for the problem.

  • 8/3/2019 Wright University Tabu Search Part1

    14/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Definition of Terms

    Tabuo Characteristic provided a move, or attributes of

    a move, that put that move off limits

    Tabu listo Memory structure of move attributes that are

    currently in off limits status

    Tabu tenureo Number of iterations for which a move, or its

    attributes, remain on the tabu list

  • 8/3/2019 Wright University Tabu Search Part1

    15/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    What is considered Tabu?

    Keep recent variables from toggling Freeze in-place recently moved nodes Freeze recently moved edges Keep variables in or out of basis

    The Tabu restriction simply prevents revisiting, orcycling which is a non-productive process. Neededto escape the local optima trap.

  • 8/3/2019 Wright University Tabu Search Part1

    16/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Definition of Terms

    Attributes of solutionso Those aspects that define a solutionoA move between solutions can be defined as a

    change to some defined set of attributes of the

    solution Attributive memory

    o Stores information about solution attributes thatare changed implementing a move

    Explicit memoryo Records the actual solution

  • 8/3/2019 Wright University Tabu Search Part1

    17/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Still More Definitions

    Aggressive explorationo Make the best move possible subject to the

    tabu restrictions in place

    Cyclingo Cycling occurs when, after a series of moves,

    the search returns to a previously visitedsolution and assumes the same searchtrajectory as previously explored

    Neighborhoodo Defined within the context of the application

  • 8/3/2019 Wright University Tabu Search Part1

    18/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Interesting Concepts

    Dynamic tabu tenureo Change the tabu tenure for any attributeo Can have each entry in tabu list have varied

    tenureso Can change the tenure based on knowledge

    gleaned from the search Called a reactive tabu search strategy

    Aspiration criteriao Means to override tabu statuso Do not want to miss good solutionso Provides additional flexibility in the search

  • 8/3/2019 Wright University Tabu Search Part1

    19/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    More on Aspiration Criteria?

    Motivation - do NOT pass on goodsolutions

    Typical criteria:oBest-so-faroBest found in some set of solutionsoBest according to some other criteria

    Multiple types of aspiration criteria arefine

  • 8/3/2019 Wright University Tabu Search Part1

    20/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Short-Term Tabu Search

    Sufficient for many problemso Hands-on will only focus on short-term

    aspects

    Tabu tenure can provide enough of anescape mechanism for an effectivesearchoConcept of non-improving moves

    forces route out of local optimality

    basinoReactively increasing the tabu tenure

    improves the escape mechanism

    The more difficult problems requireadditional tabu search constructs

  • 8/3/2019 Wright University Tabu Search Part1

    21/30

    Department of Biomedical, Human Factors, & Industrial Engineering

  • 8/3/2019 Wright University Tabu Search Part1

    22/30

    Department of Biomedical, Human Factors, & Industrial Engineering

  • 8/3/2019 Wright University Tabu Search Part1

    23/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    A Simple Example

    (use the spreadsheet provided)

  • 8/3/2019 Wright University Tabu Search Part1

    24/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Maximize:78x1+ 73x2 + 81x3 + 68x4 + 59x5 + 48x6 + 45x7 + 22x8+ 46x9 +

    96x10 + 29x11 + 56x12 + 26x13 + 74x14 + 18x15

    Subject to:

    29x1+ 33x2 + 28x3 + 7x4 + 6x5 + 27x6 + 6x7 + 1x8+32x9 + 38x10 + 31x11 + 22x12 + 27x13 + 1x14 + 9x15 153.5

    1x1+ 10x2 + 7x3 + 14x4 + 14x5 + 13x6 + 8x7 + 13x8 +

    10x9 + 7x10 + 3x11 + 11x12 + 10x13 + 8x14 + 9x15 69

    32x1+ 24x2 + 26x3 + 12x4 + 14x5 + 1x6 + 34x7 + 6x8 +

    3x9 + 18x10 + 20x11 + 15x12 + 8x13 + 7x14 + 36x15 128

    xi = 0,1

    MKP Example(should look somewhat familiar)

  • 8/3/2019 Wright University Tabu Search Part1

    25/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Consider Simple Model

  • 8/3/2019 Wright University Tabu Search Part1

    26/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Move - Toggle a variable to 1 andanother to 0

    Selected Move - Largest change, toinclude non-improving movesoAll moves will remain feasibility

    Tabu list - prohibit variable toggled to 0

    from toggling back to 1 again for 4iterations

    Simple Case One

    P bl ith Si l

  • 8/3/2019 Wright University Tabu Search Part1

    27/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Problems with SimpleCase

    Solution does not improve very much Never really comes near the constraint

    limits for feasibility A safe search, but not nearly aggressive

    enough to provide good solutions

  • 8/3/2019 Wright University Tabu Search Part1

    28/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Move Part 1 - Toggle a variable to 1o Pick largest improvement as the moveoAllow solution to become infeasibleo Consider allowing infeasibility for multiple

    steps Move Part 2 - Toggle a variable to 0

    o Continue until feasibility achievedo Drop item from most violated constraint

    whose constraint coefficient is largest

    A Better Simple Case

  • 8/3/2019 Wright University Tabu Search Part1

    29/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Better initial solutions to start thesearch

    Move and solution attributes you mightwant to track

    How might we diversify the search? What type of solutions might we

    consider saving in an elite list? Could we make random moves?

    Possible Extensions

  • 8/3/2019 Wright University Tabu Search Part1

    30/30

    Department of Biomedical, Human Factors, & Industrial Engineering

    Questions?