48
Introduction to  Artificial Intelligence Pr oblem Solv ing and St rategi es Solving problems by searching Search Str ategies Blind search Inf ormed search Game pl ay se ar ch Problem reduct ion paradi gm Search graph  AND/OR graph  No. 3

Chapt 3 Problem Solving

Embed Size (px)

Citation preview

Page 1: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 1/48

Introduction to

Artificial Intelligence

Problem Solving and Strategies

Solving problems by searchingSearch Strategies

Blind searchInformed search

Game play searchP roblem reduction paradigm

Search graph AND/OR graph

No. 3

Page 2: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 2/48

Introduction to

Artificial IntelligenceProblem as a State Space Search

To build a system to solve a particular problem, we need to:D efine the problem : must include precise specifications ~initial solution & final solution.A nalyze the problem : select the most important features thatcan have an immense impact.Isolate and represent : convert these important features intoknowledge representation.Identify problem solving technique(s) : choose the besttechnique and apply it to particular problem.

Page 3: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 3/48

Introduction to

Artificial IntelligenceThe Quest

Typical questions that need to be answered:Is the problem solver guaranteed to find a solution?W ill the system always terminate or caught in a infinite loop?If the solution is found, it is optimal ?W hat is the complexity of searching process?H ow the system be able to reduce searching complexity?H ow it can effectively utilize the representation paradigm?

Page 4: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 4/48

Introduction to

Artificial IntelligenceImportant Terms

Search space possible conditions and solutions.Initial state state where the searching processstarted.Goal state the ultimate aim of searching process.Problem space ³ what to solve´Searching strategy strategy for controlling thesearch.Search tree tree representation of search space,showing possible solutions from initial state.

Page 5: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 5/48

Introduction to

Artificial IntelligenceExample: Travelling Salesperson Problem

B

C

D

E

75

50

100

100125

125125

75

50

A

Suppose a salesperson has five cities to visit and theymust return home. Goal find the shortest path to travel.

Page 6: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 6/48

Introduction to

Artificial Intelligence

A

A A A

C

C

C

C

D

D

D

D

The possible search space for TS P .

Example: Traveling Salesperson Problem (cont)

Page 7: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 7/48

Introduction to

Artificial Intelligence

The 8 P uzzle Games contains eight tiles that can bemoved around in nine spaces.

A im : to arrange the tiles sequence using defined legalmovement (Eg: move the blank board).

1 2 3

4 6

7 8

5

Example: Eight Puzzle Games

Page 8: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 8/48

Page 9: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 9/48

Introduction to

Artificial IntelligenceSearching for Solution

Search through state space (explicitly using searching tree).

Node expansion :- generating new node related to previous nodes.

Concepts:

State :- conditions in which the node corresponds.P arent node :- the superior node

P ath cost :- the cost, from initial to goal state.

Depth:- number of steps along the path from initial state

Page 10: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 10/48

Introduction to

Artificial IntelligenceNode expansion

Page 11: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 11/48

Introduction to

Artificial IntelligenceNode expansion (initial)

Page 12: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 12/48

Introduction to

Artificial IntelligenceNode expansion (expanding A rad)

Page 13: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 13/48

Introduction to

Artificial IntelligenceNode expansion (expanding Sibiu)

Page 14: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 14/48

Introduction to

Artificial IntelligenceMeasuring Searching Performance

The output from problem-solving (searching) algorithm is either FAILURE or SOLUTION.

Four ways:

Completeness : is guaranteed to find a solution?Optimality : does it find optimal solution ?

Time complexity : how long?

Space complexity : how much memory?

Complexity : branching factor ( b ), depth ( d ), and max.depth ( m )

Page 15: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 15/48

Introduction to

Artificial IntelligenceSearching Strategies

Heuristic search searchprocess takes place bytraversing search space withapplied rules (information).

Techniques : Greedy Best FirstSearch, A* A lgorithm

There is no guarantee thatsolution is found.

Blind search traversing thesearch space until the goalnodes is found (might be doingexhaustive search).

T echniques : Breadth FirstUniform Cost , D epth first,Interactive D eepeningsearch .

Guarantees solution.

Page 16: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 16/48

Introduction to

Artificial IntelligenceBlind Search : Breadth First Search (BFS)

Strategy ~ search all the nodes expanded at given depthbefore any node at next level.

Concept : First In First Out (FIFO) queue.Complete ?: Yes with finite b (branch).

Space : similar to complexity ± keep nodes in everymemory

Optimal ? = Yes (if cost =1)

Page 17: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 17/48

Introduction to

Artificial IntelligenceBlind Search : Breadth First Search

1 2

43

Page 18: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 18/48

Introduction to

Artificial IntelligenceBlind Search : Uniform-Cost Search (UCS)

Strategy ~ search all the nodes expanded with the lowestpath cost.

If all path cost equals identical to BFS.Complete ?: Yes (if finite depth and total cost is small)

Optimality ? : Yes (proportion to cost function)

Page 19: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 19/48

Introduction to

Artificial IntelligenceBlind Search : Depth First Search (DFS)

Strategy ~ search all the nodes expanded in deepest path.

Last In First Out concept.

Complete ?: No

Optimality ? : No

Page 20: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 20/48

Introduction to

Artificial IntelligenceBlind Search : D epth First Search ( D FS)

1 2 3

4 5

««.

N+1

Page 21: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 21/48

Introduction to

Artificial IntelligenceHeuristic Search :

Important aspect: formationof heuristic function ( h(n)).

Heuristic functionadditional knowledge to

guide searching strategy(short cut).

Distance: heuristic functioncan be straight linedistance (SLD)

A*

B C*

D

E

h(n)=0

h(n)= 34

h(n)=2 4

h(n)=67

h(n)=12h(n)=9

Page 22: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 22/48

Introduction to

Artificial IntelligenceHeuristic Search : Heuristic Function

Page 23: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 23/48

Introduction to

Artificial Intelligence

Heuristic Search :Greedy-Best Search

Tries to expand the node that is closest to the goal.

Evaluates using only heuristic function : f (n) = h(n)

P ossibly lead to the solution very fast.

P roblem ? ~ can end up in sub-optimal solutions(doesn¶t take notice of the distance it travels).

Complexity and time:Complete & optimal ? : No (stuck in infinite loop)

Page 24: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 24/48

Introduction to

Artificial Intelligence

Heuristic Search :Greedy-Best Search

1

2

Page 25: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 25/48

Introduction to

Artificial Intelligence

Searching comparison (D

FS, BFS, GBFS)

A :5

CB

FED

H

G

J

:4:5

:4 :2 :6 :4

:0 :3

D iscussion:

Page 26: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 26/48

Introduction to

Artificial IntelligenceH euristic Search :Greedy-Best Search

3

Page 27: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 27/48

Introduction to

Artificial IntelligenceH euristic Search : A* Algorithm

W idely known algorithm ± (pronounced as ³ A star´search).

Evaluates nodes by combining g(n)³cost to reachthe node´ and h(n) ³ cost to get to the goal´

f (n) = g(n) + h(n), f (n) estimated cost of thecheapest solution.

Complete and optimal ~ since evaluates all paths.Time ? ~ a bit time consuming

Space ? ~ lot of it!

Page 28: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 28/48

Introduction to

Artificial Intelligence

Heuristic Search :A*

A

lgorithm

S

GE

DA

G¶ H

:10

:8 :9

:0:4 :0 :3

2 3

25 13

Path cost for S- D -G

f (S) = g(S) + h(S)

= 0 + 10 10 f (D) = (0+3) + 9 12

f (G) = (0+3+3) + 0 6

Total path cost = f (S)+ f (D)+ f (G) 28

Path cost for S- A -G¶

f (S) = 0 + 10 10 f (A) = (0+2) + 8 10 f (G¶) = (0+2+2) + 0 4

Total path cost = f (S)+ f (A)+ f (G¶) 24

* P ath S-A-G is chosen= Lowest cost

Page 29: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 29/48

Introduction to

Artificial IntelligenceHeuristic Search : A* A lgorithm

Page 30: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 30/48

Introduction to

Artificial IntelligenceHeuristic Search : A* A lgorithm

Page 31: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 31/48

Introduction to

Artificial IntelligenceHeuristic Search : A* A lgorithm

Page 32: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 32/48

Introduction to

Artificial IntelligenceHeuristic Search : A* A lgorithm

Page 33: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 33/48

Introduction to

Artificial IntelligenceHeuristic Search : A* A lgorithm

Page 34: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 34/48

Introduction to

Artificial Intelligence

Issues in Heuristic SearchSearching using heuristic function does not solely ondirected solution but the best algorithm to findshortest path towards goal.

Admissible attempt to find possible shortest pathto a goal whenever it exists.

Informedness question in what sense theheuristic function is better than another.

Monotonicity question if the best state isdiscovered by heuristic search, is there anyguarantee that the same state won¶t be found later at lowest searching cost?

Page 35: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 35/48

Introduction to

Artificial Intelligence

Games andA

I ?

Game playing is a good problem for AI research: All the information is available (human and computer have equal information.Non-trivial (requires decision making + need todisplay ³ intelligence´).Can be played in a controlled environmentCan compare humans and computers ability directly(percentage of wins/losses).For fun ?

Page 36: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 36/48

Page 37: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 37/48

Introduction to

Artificial IntelligenceGames ..

Backgammon Checkers Reversi

Page 38: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 38/48

Introduction to

Artificial IntelligenceSo what ?

³ Deep Blue´ (IBM)parallel processor, 32 nodeseach chip can search 200 million configurations/secondmemorizes starts, end-gamespower based on speed and memory: no common sense

Kasparov vs. Deep Blue, May 19976 game full-regulation chess match (sponsored by ACM)

Kasparov lost the match (2.5 to 3.5)a historic achievement for computer chess: the first time acomputer is the best chess-player on the planet

Page 39: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 39/48

Introduction to

Artificial IntelligenceSo what ?

Checkers/ D raughtscurrent world champion is Chinook,can beat any human + uses alpha-beta search

Othello (Reversi)computers can easily beat the world experts

Backgammon

system which learns is ranked in the top 3 in the worlduses neural networks to learn from playing manymany games against itself

Page 40: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 40/48

Introduction to

Artificial Intelligence

Page 41: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 41/48

Introduction to

Artificial Intelligence

8 8 P P u u z z

z z l l ee

Initial State

Goal State

Solution:

a p c p f p j p l p m

Search Space

Page 42: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 42/48

Introduction to

Artificial Intelligence8 8- -puzzle : puzzle : RulesRules

Start

Goal

Page 43: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 43/48

Introduction to

Artificial Intelligence

H euristic I:Number of tiles out of place

Current State Goal State

8 8- -puzzle : puzzle : HeuristicsHeuristics

Number of tiles out place: 5 @ h( n) = 5

Page 44: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 44/48

Introduction to

Artificial Intelligence

Depth-First SearchBreath-First SearchBest-First Search

A* Algorithm

8 8- -puzzle : puzzle : StrategiesStrategies

Page 45: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 45/48

Introduction to

Artificial Intelligence

H euristic II:Sum of distances out of place

Current State Goal State

8 8- -puzzle : puzzle : HeuristicsHeuristics

Tile D istances out of place

2 1

8 2

1 1

Sum of distances out of place: 4@

h(n)=

4

Page 46: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 46/48

Introduction to

Artificial Intelligence

Goal State

H euristic

8 8- -puzzle : puzzle : HeuristicsHeuristics

Page 47: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 47/48

Introduction to

Artificial Intelligence

Exercise:

2 8 3

7 51 6 4

1 2 3

8 47 6 5Goal StateInitial State

H euristic: sum of distances out of place

Consider the following 8-p uzzle problem:

Construct a state space for the above problem usingGreedy Best-First strategy

Page 48: Chapt 3 Problem Solving

8/7/2019 Chapt 3 Problem Solving

http://slidepdf.com/reader/full/chapt-3-problem-solving 48/48

Introduction to

Artificial Intelligence

References

Cawsey, A. (1998). The Essence of ArtificialIntelligence, P rentice H all.Russell, S. and Norvig, P . (2003). Artificial

Intelligence: A Modern Approach, P rentice- H all2nd Edition.