Introduction to
Artificial Intelligent (AI)
2. Problem solving by
Searching
Define a problem
A problem can be defined by 5 components:
• Initial state
• Possible Actions: a set of actions that can be executed in
each state
• Transition (successor) function: returns the state that
results from doing an action, but does not tell which action to
use
• Goal test: check if the current state is goal state
• Cost: a cost function that reflects its own performance
measure
2
Define a problem
Problem: find path from Arad to Bucharest
3
Define a problem
Problem: 8 - Puzzle
4
Define a problem
Problem: 8 - Puzzle
How many states 8-puzzle has?
9!/2 = 181,440 states
Transition (successor) function returns
state/alternatives.
5
Define a problem
Problem: 15 - Puzzle
In 1878, Sam Loyd offered $1,000 to the first
Sam Loyd (1841-1911)
person who would solve the following problem America’s greatest
puzzle-expert
But no one ever won the prize!!
6
Define a problem
Problem: 15 - Puzzle
Search methods are
used to explore and
find the optimal
alternatives
15-puzzle has 16!/2 ~ 1.04 x 1013 states
→ too large to store and would take too
long to learn
7
Define a problem as a search
problem
Search problem’s components:
• Initial state
• State space S
• Successor function:
x ∈ S → SUCCESSORS(x) ∈ 2S
• Goal test: x ∈ S → GOAL?(x) =T or F
• Arc Cost
8
State Graph
• Each state is represent by a
node
• An arc connects a node s to
node s’ if s’ ∈ SUCCESSORS(s)
• A solution is a path
connecting the initial node to a
goal node
• The cost is the sum of the arc
costs along a path
• An optimal solution is a
solution path of minimum cost
• There might be no solution !
9
Problem formulation
A problem can be formulated in different ways.
Problem: path finding
10
#1 #2 #3
Problem formulation
#1 #2 #3
11
State space
How big is the state space of the (n2 - 1)-puzzle?
• 8-puzzle → 9! = 362,880 states 0.036s
• 15-puzzle → 16! ~ 2.09 x 1013 states 55 hours
• 24-puzzle → 25! ~ 1025 states 109 years
If a computer can do
100 million state/sec
It is often not feasible (or too expensive) to build a complete
state graph
A problem solver must construct a solution by exploring a
small portion of the graph
12
Searching the State Space
13
Search Algorithms
14
a G
b c
Blind search e
d f
Breath First Search S h
Strategy: expand a shallowest p r
q
node first
S
d e p
Search
b c e h r q
Tiers
15
a a h r p q f
p q f q c G
q c a
G
a
Blind search
Breath First Search
Strategy: expand a shallowest node first
Visited
S
Queue
1 2
First Last
First In First Out
(FIFO)
Blind search
Breath First Search
Strategy: expand a shallowest node first
Visited
S 1
Queue
2 3
First Last
First In First Out
(FIFO)
Blind search
Breath First Search
Strategy: expand a shallowest node first
Visited
S 1 2
Queue
3 4
First Last
First In First Out
(FIFO)
Blind search
Breath First Search
Strategy: expand a shallowest node first
Visited
S 1 2
Queue
4 5
First Last
First In First Out
(FIFO)
Blind search
Breath First Search
Strategy: expand a shallowest node first
Visited
S 1 2 3 4
Queue
5
First Last
First In First Out
(FIFO)
Blind search
Breath First Search
Strategy: expand a shallowest node first
Visited
S 1 2 3 4 5
Queue
Shortest Path
S 2 3 5
Blind search
Breath First Search
Strategy: expand a shallowest node first
Blind search
Breath First Search
Strategy: expand a shallowest node first
Exercise: Rat in a Maze Problem
a G
Blind search b c
Depth First Search e
d f
Strategy: expand a deepest S h
node first
p q r
S
d e p
b c e h r q
a a h r p q f
p q f q c G
q c a
G
a
2 a G
b 8 c
Blind search 1
2 2
3 e
d 9 2 f
Uniform Cost Search S h 8
1
Strategy: expand a cheapest node first: 1 p r
q
15
S 0
d 3 e 9 p 1
b 4 c e 5 h 17 r 11 q 16
11
25
a 6 a h 13 r 7 p q f
Cost
contours q c
p q f 8 G
q c a
11 G 10
a
Heuristic search
Heuristic function h(n) is a function used to estimates how close a
state is to the goal. h(n) is always positive
Heuristic search find node/path based on their heuristic value h(n).
Heuristic search uses 2 lists:
- OPEN: nodes which have yet been expanded
- CLOSED: nodes which have already expanded
On each iteration, node n with the lowest heuristic value is
expanded and generates all its successors, after that n is placed to the
closed list.
2 main heuristic search algorithms:
- Greedy search
- A* search
26
Heuristic search
Greedy Search : selects the path which appears best at that moment by
combination of DFS and BFS.
h(x) 27
Heuristic search
Greedy Search
28
Heuristic search
Greedy Search
10
5
11.2
29
Heuristic search
A* Search : find the shortest path by combine heuristic cost and cost to
reach node n from start state
Uniform-cost Greedy search
search
30
Heuristic search
A* Search : find the shortest path by combine heuristic cost and cost to
reach node n from start state
http://theory.stanford.edu/~amitp/GameProgramming/Heuristics.html 31
Heuristic search
Greedy
A*
Uniform Cost
32
Heuristic search
A* Search
33
