CS 1571 Introduction to AI

Lecture 3

Problem solving by searching

Milos Hauskrecht
[email protected]
5329 Sennott Square

CS 1571 Intro to AI M. Hauskrecht

A search problem
Many interesting problems in science and engineering are solved
using search
A search problem is defined by:
• A search space:
– The set of objects among which we search for the solution
Examples: routes between cities, or n-queens configuration
• A goal condition
– Characteristics of the object we want to find in the search
space?
– Examples:
• Path between cities A and B
• Non-attacking n-queen configuration

CS 1571 Intro to AI M. Hauskrecht

1
 Graph representation of a search problem
• Search problems can be often represented using graphs
• Typical example: Route finding
– Map corresponds to the graph, nodes to cities, links valid
moves via available connections
– Goal: find a route (sequence of moves) in the graph from S
to T
G H J
B
S C
K
start I

A D
L
E T target
F
CS 1571 Intro to AI M. Hauskrecht

Graph search
• Less obvious conversion:
Puzzle 8. Find a sequence of moves from the initial
configuration to the goal configuration.
– nodes corresponds to states of the game,
– links to valid moves made by the player

start
target

CS 1571 Intro to AI M. Hauskrecht

2
 Graph Search Problems
Search problems can be often represented as graph search
problems:
• Initial state
– State (configuration) we start to search from (e.g. start city,
initial game position)
• Operators:
– Transform one state to another (e.g. valid connections
between cities, valid moves in Puzzle 8)
• Goal condition:
– Defines the target state (destination, winning position)

Search space is now defined indirectly through:

The initial state + Operators

CS 1571 Intro to AI M. Hauskrecht

N-queens
Some problems are easy to convert to the graph search problems
• But some problems are harder and less intuitive
– Take e.g. N-queens problem.

Goal configuration

• Problem:
– We look for a configuration, not a sequence of moves
– No distinguished initial state, no operators (moves)
CS 1571 Intro to AI M. Hauskrecht

3
 N-queens
How to choose the search space for N-queens?
• Ideas? Search space:
– all configurations of N queens on the board

…..
• Can we convert it to a graph search problem?
• We need states, operators, initial state and goal condition.

initial goal
…
CS 1571 Intro to AI M. Hauskrecht

N-queens
Search space:
– all configurations of N queens on the board

• To convert it to a graph search problem we need states,

operators, initial state and goal condition.

goals
initial
…

Initial state: an arbitrary N-queen configuration

Operators (moves): change a position of one queen
CS 1571 Intro to AI M. Hauskrecht

4
 N-queens
An alternative way to formulate the N-queens problem as a
search problem:
• Search space: configurations of 0,1,2, … N queens
• Graph search:
– States configurations of 0,1,2,…N queens
– Operators: additions of a queen to the board
– Initial state: no queens on the board

start

CS 1571 Intro to AI M. Hauskrecht

Graph search
N-queens problems
• This is a different graph search problem when compared to
Puzzle 8 or Route planning:
We want to find only the target configuration, not a path

CS 1571 Intro to AI M. Hauskrecht

5
 Two types of graph search problems
• Path search
– Find a path between states S and T
– Example: traveler problem, Puzzle 8
– Additional goal criterion: minimum length (cost) path

• Configuration search (constraint satisfaction search)

– Find a state (configuration) satisfying the goal condition
– Example: n-queens problem
– Additional goal criterion: “soft” preferences for
configurations, e.g. minimum cost design

CS 1571 Intro to AI M. Hauskrecht

Search
• Search (process)
– The process of exploration of the search space
• The efficiency of the search depends on:
– The search space and its size
– Method used to explore (traverse) the search space
– Condition to test the satisfaction of the search objective
(what it takes to determine I found the desired goal object

CS 1571 Intro to AI M. Hauskrecht

6
 Comparison of two problem formulations
Solution 1:

Operators: switch one of the queens

 16 
  - all configurations
 4

Solution 2:

Operators: add a queen to the leftmost unoccupied column

1  4  42  43  44  45 - configurations altogether
CS 1571 Intro to AI M. Hauskrecht

Even better solution to the N-queens

Solution 2:

Operators: add a queen to the leftmost unoccupied column

 4 5 - configurations altogether
Improved solution with a smaller search space
Operators: add a queen to the leftmost unoccupied column
such that it does not attack already placed queens

 1  4  4 * 3  4 * 3 * 2  4 * 3 * 2 * 1  65
- configurations altogether

CS 1571 Intro to AI M. Hauskrecht

7
 Search
• Search (process)
– The process of exploration of the search space
• The efficiency of the search depends on:
– The search space and its size
– Method used to explore (traverse) the search space
– Condition to test the satisfaction of the search objective
(what it takes to determine I found the desired goal object

Think twice before solving the problem:

Choose the search space wisely
CS 1571 Intro to AI M. Hauskrecht

Search
• Search (process)
– The process of exploration of the search space
• The efficiency of the search depends on:
– The search space and its size
– Method used to explore (traverse) the search space
– Condition to test the satisfaction of the search objective
(what it takes to determine I found the desired goal object

CS 1571 Intro to AI M. Hauskrecht

8
 Search process
Exploration of the state space through successive application of
operators from the initial state
• Search tree = structure representing the exploration trace
– Is built on-line during the search process
– Branches correspond to explored paths, and leaf nodes to
the exploration fringe
Arad

Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu

Vilcea

CS 1571 Intro to AI M. Hauskrecht

Search tree
• A search tree = (search) exploration trace
– different from the graph representation of the problem
– states can repeat in the search tree

Graph

Arad Search tree

built by exploring paths
starting from city Arad
Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu

Vilcea
CS 1571 Intro to AI M. Hauskrecht

9
 Search tree

A branch in the search tree

Arad
= a path in the graph

Zerind Sibiu Timisoa

ra

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea
CS 1571 Intro to AI M. Hauskrecht

Search tree

A branch in the search tree

Arad
= a path in the graph

Zerind Sibiu Timisoa

ra

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea
CS 1571 Intro to AI M. Hauskrecht

10
 General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

CS 1571 Intro to AI M. Hauskrecht

General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad

CS 1571 Intro to AI M. Hauskrecht

11
 General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad Node chosen to be expanded next

CS 1571 Intro to AI M. Hauskrecht

General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad Check if the node satisfied the goal

CS 1571 Intro to AI M. Hauskrecht

12
 General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad Expanded node

Zerind Sibiu Timisoara

Generated (or active, or open) nodes

CS 1571 Intro to AI M. Hauskrecht

General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad
Newly selected node

Zerind Sibiu Timisoara

CS 1571 Intro to AI M. Hauskrecht

13
 General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad
Check if it is the goal

Zerind Sibiu Timisoara

CS 1571 Intro to AI M. Hauskrecht

General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad
Expanded nodes

Zerind Sibiu Timisoara

Generated (active, open) nodes

Arad Oradea

CS 1571 Intro to AI M. Hauskrecht

14
 General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad

Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu

Vilcea
CS 1571 Intro to AI M. Hauskrecht

General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

Arad

Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea
CS 1571 Intro to AI M. Hauskrecht

15
 General search algorithm

General-search (problem, strategy)

initialize the search tree with the initial state of problem
loop
if there are no candidate states to explore next return failure
choose a leaf node of the tree to expand next according to a strategy
if the node satisfies the goal condition return the solution
expand the node and add all of its successors to the tree
end loop

• Search methods differ in how they explore the space, that

is how they choose the node to expand next !!!!!

CS 1571 Intro to AI M. Hauskrecht

Implementation of search
• Search methods can be implemented using the queue structure

General search (problem, Queuing-fn)

nodes  Make-queue(Make-node(Initial-state(problem)))
loop
if nodes is empty then return failure
node  Remove-node(nodes)
if Goal-test(problem) applied to State(node) is satisfied then return node
nodes  Queuing-fn(nodes, Expand(node, Operators(node)))
end loop

• Candidates are added to the queue structure (named nodes)

• Queuing function determines what node will be selected next
CS 1571 Intro to AI M. Hauskrecht

16
 Implementation of search
• A search tree node is a data-structure that is a part of the
search tree
parent
ST
State Other attributes:
Node - state value (cost)
- depth
state - path cost

children

• Expand function – applies Operators to the state represented

by the search tree node. Together with Queuing-fn it fills the
attributes.

CS 1571 Intro to AI M. Hauskrecht

Search
• Search (process)
– The process of exploration of the search space
• The efficiency of the search depends on:
– The search space and its size
– Method used to explore (traverse) the search space
– Condition to test the satisfaction of the search objective
(what it takes to determine I found the desired goal object

CS 1571 Intro to AI M. Hauskrecht

17
 Uninformed search methods
• Search techniques that rely only on the information available
in the problem definition
– Breadth first search
– Depth first search
– Iterative deepening
– Bi-directional search

For the minimum cost path problem:

– Uniform cost search

CS 1571 Intro to AI M. Hauskrecht

Search methods
Properties of search methods :
• Completeness.
– Does the method find the solution if it exists?

• Optimality.
– Is the solution returned by the algorithm optimal? Does it
give a minimum length path?

• Space and time complexity.

– How much time it takes to find the solution?
– How much memory is needed to do this?

CS 1571 Intro to AI M. Hauskrecht

18
 Parameters to measure complexities.
• Space and time complexity.
– Complexity is measured in terms of the following tree
parameters:
• b – maximum branching factor
• d – depth of the optimal solution
• m – maximum depth of the state space
Branching factor
The number of
applicable operators

CS 1571 Intro to AI M. Hauskrecht

Breadth first search (BFS)

• The shallowest node is expanded first

CS 1571 Intro to AI M. Hauskrecht

19
 Breadth-first search
• Expand the shallowest node first
• Implementation: put successors to the end of the queue (FIFO)

queue Arad
Arad

Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea

CS 1571 Intro to AI M. Hauskrecht

Breadth-first search

queue Zerind
Sibiu
Timisoara
Arad

Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea

CS 1571 Intro to AI M. Hauskrecht

20
 Breadth-first search

queue Sibiu
Timisoara
Arad
Oradea
Arad

Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea

CS 1571 Intro to AI M. Hauskrecht

Breadth-first search

queue Timisoara
Arad
Oradea
Arad
Arad
Oradea
Fagaras
Romnicu Vilcea
Zerind Sibiu Timisoara

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea

CS 1571 Intro to AI M. Hauskrecht

21
 Breadth-first search

queue Arad
Oradea
Arad
Oradea
Arad Fagaras
Romnicu Vilcea
Arad
Zerind Sibiu Timisoara
Lugoj

Arad Oradea Arad Oradea Fagaras Rimnicu Arad Lugoj

Vilcea

CS 1571 Intro to AI M. Hauskrecht

Properties of breadth-first search

• Completeness: Yes. The solution is reached if it exists.

• Optimality: Yes, for the shortest path.

• Time complexity: ?

• Memory (space) complexity: ?

CS 1571 Intro to AI M. Hauskrecht

22
 BFS – time complexity
depth number of nodes

b 0 1

1 21=2
d
2 22=4

3 23=8

d 2d (bd )

d+1 2d+1 (bd+1)

Expanded nodes: O (b d ) Total nodes: O ( b d 1 )

CS 1571 Intro to AI M. Hauskrecht

Properties of breadth-first search

• Completeness: Yes. The solution is reached if it exists.

• Optimality: Yes, for the shortest path.

• Time complexity:
1  b  b 2    b d  O (b d )
exponential in the depth of the solution d

• Memory (space) complexity: ?

CS 1571 Intro to AI M. Hauskrecht

23