DongChul Kim, HwangRyol Ryu _ TSP Slides

download DongChul Kim, HwangRyol Ryu _ TSP Slides

of 18

Transcript of DongChul Kim, HwangRyol Ryu _ TSP Slides

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    1/18

    Traveling Salesman Problem

    DongChul KimHwangRyol Ryu

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    2/18

    Introduction

    Research Goal

    What you will learn

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    3/18

    What Is TSP?

    Shortest Hamiltonian cycle (i.e. tour).

    Grow exponentially

    Current Definition of TSP:Given a number of cities and the costs of

    traveling from one to the other, what is the

    cheapest round trip route that visits each city

    and then returns to the starting city?

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    4/18

    History of TSP

    Irish mathematician Sir William Rowan

    Hamilton and the British mathematician

    Thomas Penyngton Kirkman

    Hamiltons Iconsian game

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    5/18

    History of TSP (1)

    The general form of TSP appeared in 1930sby Karl Menger in Vienna and Havard.

    A breakthrough by George Dantzig, Ray

    Fulkerson, and Selmer Johnson in1954. 49 - 120 550 - 2,392 - 7,397 19,509 cities

    From year 1954 to year 2001.

    24,098 cities by David Applegate, RobertBixby, Vasek Chvatal, William Cook, andKeld Helsgaun in May 2004.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    6/18

    Branch & Lower Bound

    An algorithmic technique to find the optimal

    solution by keeping the best solution found so

    far.

    Standard to measure performance of TSP

    heuristics.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    7/18

    2.0 TSP Approximation Algorithm

    Double Minimum Spanning Tree

    Return a tour of length at most twice the shortest

    tour.

    Algorithm:

    1. Construct the minimal spanning tree

    2. Duplicate all its edges. This gives us an Euler

    cycle.

    3. Traverse the cycle, but do not visit any node

    more than once, taking shortcuts when it passes a

    visited node.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    8/18

    2.0 TSP Approximation Algorithm (2)

    2.0? TSP

    2.0 is TSP version number?

    Tour of length is at most twice the length ofMST.

    MST < Euler Cycle = 2 * MST

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    9/18

    1.5 TSP Approximation Algorithm(Known as Christofides Heuristics)

    Professor Nicos Christofides extended the 2.0 TSP

    and published that the worst-case ratio of the

    extended algorithm was 3/2.

    Algorithm:1. Compute MST graph T.

    2. Compute a minimum-weighted matching graph M.

    3. Combine T and M as edge set and Compute an

    Euler Cycle.

    4. Traverse each vertex taking shortcuts to avoid

    visited nodes.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    10/18

    1.5 TSP Approximation Algorithm (2)(Known as Christofides Heuristics)

    What is a Minimum-weighted Matching?

    It creates a MWM on a set of the nodes

    having an odd degree.

    Why odd degree?

    Property of Euler Cycle

    Why 1.5 TSP?

    MST < Euler Cycle = MWM+MST

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    11/18

    1.5 TSP Approximation Algorithm (3)(Known as Christofides Heuristics)

    Minimum-weighted Matching example

    MWM = MST

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    12/18

    Matching Algorithm

    Smile Matching Algorithm Bad matching

    Better matching

    Fixed Bad matching problem.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    13/18

    Matching Algorithm (2)

    Improved Smile Matching Algorithm

    1. Choose the two nodes in the farthest

    distance

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    14/18

    Matching Algorithm (3)

    Improved Smile Matching Algorithm

    2. Each end node is connected to the node in

    the closest distance.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    15/18

    PTAS Algorithm

    (Polynomial Time Approximation Sch

    eme) The status of Euclidean TSP remained open.

    PTAS = Polynomial time algorithm, for each

    c > 1, can approximate the problem within a

    factor 1 + 1/c.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    16/18

    PTAS Algorithm (2)

    The central idea of the PTAS is that the plane

    can be recursively partitioned and by using a

    dynamic programming on Quadtree, it finds

    an optimal tour.

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    17/18

    Other approximation schemes

    Minimum Steiner Tree

    K-TSP and K-MST

    Min CostP

    erfect Matching

  • 8/6/2019 DongChul Kim, HwangRyol Ryu _ TSP Slides

    18/18

    Demonstration