Previous Search Problems
• Search-Based Problem Solving:
– Focuses on navigating a state space, where each state may or may not represent a partial
or complete solution.
– Heuristics and search strategies are used to find paths leading to a goal state.
• Constraint Satisfaction Problem:
– Involves systematically assigning values to variables while satisfying all constraints.
– Constraints guide the process to prune inconsistent paths early, making the search space
more structured and manageable.
Constraint Satisfaction Problem
• A constraint satisfaction problem (or CSP) has three components V, D, C
– V is defined by a set of variables, X1, X2, …, Xn
– C is a set of constraints, C1, C2, …, Cm
– Each variable Xi has a nonempty domain Di of possible values.
– Each constraint Ci consists of a pair <scope, rel>, where scope is a tuple of
variables and rel is the relation, either represented explicitly or abstractly.
– X1 and X2 both have the domain {A, B}
– <(X1, X2), X1 ≠ X2>
– <(X1, X2), (A, B)>
Constraint Satisfaction Problem
• Each state in a CSP is defined by an assignment of values to some or all of the
variables.
• An assignment that does not violate any constraints is called a consistent or
legal assignment.
• A complete assignment is one in which every variable is assigned.
• A solution to a CSP is consistent and complete assignment.
• Allows useful general-purpose algorithms with more power than standard
search algorithms.
• V = {WA, NT, SA, Q, NSW, V, T}
• D = {red, green, blue}
• C = {adjacent nodes should not have
same color}
– E.g. WA ≠ NT
– E.g: (WA, NT) = {(red, green), (red,
blue), (green, red), …}
• Constraint graph: Nodes are
variables, arcs are constraints
• Binary CSP: Each constraint relates
two variables.
Types of Constraint
• Unary constraints involve a single variable,
– e.g., SA ≠ green
• Binary constraints involve pairs of variables,
– e.g., SA ≠ WA
• Higher-order constraints involve 3 or more variables
Backtracking Search for CSPs
Improving Backtracking Efficiency
• General-purpose methods can give huge gains in speed:
– Which variable should be assigned next?
– In what order should its values be tried?
– Can we detect inevitable failure early?
Minimum remaining value
• Choose the variable with the fewest legal values
Most Constraining Variable
• Choose the variable with the most constraints on remaining variables.
Least Constraining Value
• Given a variable to assign, choose the least constraining value:
– The one that rules out the fewest values in the remaining variables
Forward Checking (Filtering)
• Idea
– Keep track of remaining legal values for unassigned variables
– Terminate search when any variable has no legal values
Map Coloring (Forward Checking / Filtering)
WA NT Q NSW V SA T
Initial domains RGB RGB RGB RGB RGB RGB RGB
After WA = red R GB RGB RGB RGB GB RGB
After Q = green R B G RB RGB B RGB
After V = blue R B G R B RGB
After V = G R B G R G
Cryptarithmetic Problem
• Type of Constraint Satisfaction
S E N D
Problem
+ M O R E
• Constraints
M O N E Y
– No two letters have same value
– Sum of digits must be as shown in
problem 9 5 6 7
– No leading zeros are allowed + 1 0 8 5
– Digits that can be assigned to a
1 0 6 5 2
word/alphabet (0-9) range
N-Queen Problem
• Type of Constraint Satisfactory Problem in AI
• ‘NxN’ square grid board
• ‘N’ Queens needs to be placed on them
• Following constraints needs to be satisfied
– No row should contain more than one queen
– No column should contain more than one queen
– No diagonal should contain more than one queen
– There should be no row/column without any queen
4-Queen Problem (Backtracking)
Q1 Q1 Q1 Q1
………………
………………
Adversarial Search
• Adversarial search in AI is a type of search strategy used to make optimal
decisions in situations where there is an opponent or adversary, such as in two-
player games (e.g., chess, checkers).
• The goal is to find the best possible move for a player while considering that
the opponent is trying to counteract and minimize that player's success.
• Perfect decision (e.g. Chess, Tic-Tac-Toe)
• Imperfect decision (e.g. Card game)
Minimax Search Algorithm
• Consider the following two
MAX
player game tree in which the
static scores are given from the
MIN
first players point of view.
• Apply the Minimax search
MAX
algorithm and compute the value
of the root of the tree.
• Also find the most convenient
10 9 14 18 5 4 50 3
path for Max Node.
Minimax Search Algorithm
• Generate the whole game tree to leaves
• Apply utility (payoff) function to leaves 10 MAX
• Uses DFS for expanding the tree
10 5 MIN
• Back-up values from leaves toward the
root:
– a Max node computes the maximum value
14
18 5 50 MAX
10
from its child values
– a Min node computes the minimum value
from its child values
• When value reaches the root: optimal
10 9 14 18 5 4 50 3
move is determined.
Alpha Beta Pruning Algorithm
• The problem with minimax search is that the number of game states it has to
examine is exponential in the number of moves.
Alpha Beta Pruning Algorithm
• Alpha(α) – Beta(β) proposes to find the optimal
path without looking at every node in the game 10 MAX(α)
tree.
• Max contains Alpha(α) and Min contains 10 5 MIN(β)
Beta(β) bound during the calculation.
• In both Min and Max node, we return when α
>=β which compares with its parent node only. 10 14 5 MAX(α)
• Both minimax and Alpha(α) – Beta(β) cut-off
give same path.
• Alpha(α) – Beta(β) gives optimal solution as it
10 9 14 18 5 4 50 3
takes less time to get the value for the root node.
Exercise
