4 heuristics search v2

Upload: crystalcry

Post on 01-Jun-2018

225 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/9/2019 4 Heuristics Search V2

    1/29

    Informed search algorithms

    Chapter 4

  • 8/9/2019 4 Heuristics Search V2

    2/29

    Material

    Chapter 4 Section 1 3

    Exclude memory-ounded heuristic

    search

  • 8/9/2019 4 Heuristics Search V2

    3/29

    !e"ie#$ %ree search

    & search strategy is defined y pic'ing theorder of node expansion

  • 8/9/2019 4 Heuristics Search V2

    4/29

    (est-first search

    Idea$ use an e"aluation functionf(n) for each node estimate of )desiraility*

    Implementation$

    +rder the nodes in fringe in decreasing order of desiraility

    Special cases$ greedy est-first search

    &,search

    Expand most desirale unexpanded node

  • 8/9/2019 4 Heuristics Search V2

    5/29

    !omania #ith step costs in 'm

  • 8/9/2019 4 Heuristics Search V2

    6/29

    .reedy est-first search

    E"aluation function f(n) = h(n) /heuristic0

    estimate of cost from nto goal

    e2g2 hSLD(n) straight-line distance from nto (ucharest

    .reedy est-first search expands the node

    that appearsto e closest to goal

  • 8/9/2019 4 Heuristics Search V2

    7/29

    .reedy est-first searchexample

  • 8/9/2019 4 Heuristics Search V2

    8/29

    .reedy est-first searchexample

  • 8/9/2019 4 Heuristics Search V2

    9/29

    .reedy est-first searchexample

  • 8/9/2019 4 Heuristics Search V2

    10/29

    .reedy est-first searchexample

  • 8/9/2019 4 Heuristics Search V2

    11/29

    roperties of greedy est-firstsearch

    Complete56o can get stuc' in loopse2g2 Iasi6eamtIasi6eamt

    %ime5 +/m0 ut a good heuristic cangi"e dramatic impro"ement

    Space5 +/m0 -- 'eeps all nodes inmemory

    +ptimal5 6o

  • 8/9/2019 4 Heuristics Search V2

    12/29

    &,search

    Idea$ a"oid expanding paths that arealready expensi"e

    E"aluation function f(n) = g(n) + h(n)

    g(n) cost so far to reach n

    h(n) estimated cost from nto goal

    f(n) estimated total cost of path throughnto goal

  • 8/9/2019 4 Heuristics Search V2

    13/29

    &,search example

  • 8/9/2019 4 Heuristics Search V2

    14/29

    &,search example

  • 8/9/2019 4 Heuristics Search V2

    15/29

    &,search example

  • 8/9/2019 4 Heuristics Search V2

    16/29

    &,search example

  • 8/9/2019 4 Heuristics Search V2

    17/29

    &,search example

  • 8/9/2019 4 Heuristics Search V2

    18/29

    &,search example

  • 8/9/2019 4 Heuristics Search V2

    19/29

    +ptimality of &,

    &,expands nodes in order of increasing f"alue

    .radually adds )f-contours) of nodes

    Contour ihas all nodes #ith f=fi #here fi< fi+1

  • 8/9/2019 4 Heuristics Search V2

    20/29

  • 8/9/2019 4 Heuristics Search V2

    21/29

    ;ocal search algorithms

    In many optimi

  • 8/9/2019 4 Heuristics Search V2

    22/29

    Example$ n-?ueens

    ut n?ueens on an n noard #ith not#o ?ueens on the same ro# column ordiagonal

  • 8/9/2019 4 Heuristics Search V2

    23/29

    @ill-climing search

    );i'e climing E"erest in thic' fog #ithamnesia)

  • 8/9/2019 4 Heuristics Search V2

    24/29

    @ill-climing search

    rolem$ depending on initial state canget stuc' in local maxima

  • 8/9/2019 4 Heuristics Search V2

    25/29

    @ill-climing search$ A-?ueens prolem

    h numer of pairs of ?ueens that are attac'ing each other either directlyor indirectly

    h = 17for the ao"e state

  • 8/9/2019 4 Heuristics Search V2

    26/29

    @ill-climing search$ A-?ueens prolem

    & local minimum #ith h = 1

  • 8/9/2019 4 Heuristics Search V2

    27/29

    ;ocal eam search

    :eep trac' of kstates rather than Bust one

    Start #ith ' randomly generated states

    &t each iteration all the successors of all 'states are generated

    If any one is a goal state stop= else selectthe ' est successors from the completelist and repeat2

  • 8/9/2019 4 Heuristics Search V2

    28/29

    .enetic algorithms

    & successor state is generated y comining t#o parentstates

    Start #ith krandomly generated states /population0 & state is represented as a string o"er a finite alphaet

    /often a string of s and 1s0 E"aluation function /fitness function02 @igher "alues foretter states2

    roduce the next generation of states y selectioncrosso"er and mutation

  • 8/9/2019 4 Heuristics Search V2

    29/29

    .enetic algorithms

    >itness function$ numer of non-attac'ing pairs of ?ueens /min max A D FG GA0

    G4F/G4HG3HGH110 31 G3F/G4HG3HGH110 GJ etc