Transcript Lecture 11-12 - เว็บไซต์บุคลากรภาควิชาวิทยาการคอมพิวเตอร์

Chapter 3
Heuristic Search Techniques (continue)
323-670 Artificial Intelligence
ดร.วิภาดา เวทย์ ประสิทธิ์ ภาควิชาวิทยาการคอมพิวเตอร์ คณะวิทยาศาสตร์ มหาวิทยาลัยสงขลานครินทร์
Constraint Satisfaction
 limit on time, cost, materials
 reduce search space
 process
 step 1. constraint are discover and propagate as
far as possible
 if no solution found  search again
 step 2. Guess about something is made to set a
new constraint and propagate with the new
constraint
 EX:
323-670 Artificial Intelligence
N=E+1
N=3
E=2
..................(1)
..................(2)
..................(3)
Lecture 11-12
Page 2
Constraint Satisfaction
 To terminate the current path:
 1. found contradiction
 If we found contradiction  then we know
that there is no solution
 2. propagation run out of stream or no further
changes that can be made on the basis of current
basis.
 backtracking : will be use to find a new path
when found contradiction in the current
path or no further change in the current
path..
 more powerful rules  less guessing needed

323-670 Artificial Intelligence
Lecture 11-12
Page 3
Constraint Satisfaction
 constraint :
 1) simple : list of possible value for a single
object  dynamic value
 describe explicitly in each
problem state
 2) complex : describe relationship between
or among objects
 dynamic value or
 static value : ex. waltz problem
p.374 ***
 DDB : dependency–directed backtracking
p.211 ***
323-670 Artificial Intelligence
Lecture 11-12
Page 4
Constraint Satisfaction Algorithm
p. 90
323-670 Artificial Intelligence
Lecture 11-12
Page 5
Cryptarithmetic Problem
323-670 Artificial Intelligence
Lecture 11-12
Page 6
Means-Ends Analysis
 Allow mixed strategy : forward chaining and
backward chaining
 Operator subgoal/ priority level for each
subgoal
 recursive process
 GPS : general problem solver [Newell and
Simon 1963]
 not suitable for very large problem
 more example of clock world is in chapter
13, p. 332
323-670 Artificial Intelligence
Lecture 11-12
Page 7
Means-Ends Analysis
Rules consist of : figure 3.15
 Left side : PRECONDITION
 Right side: describe problem state that will be changed
by the application of the rule.
323-670 Artificial Intelligence
Lecture 11-12
Page 8
Means-Ends Analysis
 PROBLEM TO SOLVE : ROBOT
Moving a desk with two things on it from one
room to another.......
 they may be more than one operator to do
the job for reducing the difference
 carry : small object
 push : large object
 change size of object operator : saw_apart ?
323-670 Artificial Intelligence
Lecture 11-12
Page 9
Means-Ends Analysis
Difference Table : figure 3.16
323-670 Artificial Intelligence
Lecture 11-12
Page 10
Means-Ends Analysis
chair
C to E = ?
chair
C to E = WALK PICKUP CARRY
323-670 Artificial Intelligence
Lecture 11-12
Page 11
Means-Ends Analysis
p. 97
323-670 Artificial Intelligence
Lecture 11-12
Page 12
Summary
The difference among the proposed algorithms
 How, at each state of the search process, a state is
selected for expansion.
 How operator to be apply to that nose is selected.
 Whether an optimal solution can be guaranteed.
 Whether a given state may end up being considered
more than once.
 How many search descriptions must be maintained
throughout the search process.
 Under what circumstances should a particular search
path be abandoned.
323-670 Artificial Intelligence
Lecture 11-12
Page 13
END Chapter 3
Jim Miller
323-670 Artificial Intelligence
Lecture 11-12
Page 14