Artificial Intelligence

Outline
 Last Class
 Beyond Classical Search
 Types of Problems
 Optimization problems
 Local Search Algorithms
 Hill Climbing
 Hill climbing features
 Hill climbing Problems
 Hill Climbing variants

 Today
 Adversarial Search
 Formalization of the problem
 Game Tree
 Mini Max Algorithm
 Alpha-beta Pruning
 Single Agent environment
 Until now, we have discussed single agent environment
 only one person or agent searching the solution space to find
the goal or the solution.
 often expressed in the form of a sequence of actions

 There might be some situations where more than one

agent/person is searching for the solution in the same
search space
 usually occurs in game playing where two opponents
(adversaries) are searching for a goal.
 Multi-agent environment
 Multi-agent environment
 The environment with more than one agent
 Each agent is an opponent of other agent and playing
against each other
 Each agent needs to consider the action of other agent
and effect of that action on their performance.
 Adversarial Search
 Adversarial Search
 Searches in which two or more players with conflicting
goals are trying to explore the same search space for the
solution
 often known as Games

 For example, in a game of tic-tac-toe player one might

want that he should complete a line with crosses while
at the same time player two wants to complete a line
of zeros
 Types of Games in AI

Deterministic Chance Moves

Perfect information Chess, Checkers, go, Backgammon,

Othello monopoly

Imperfect information Battleships, blind, tic-tac- Bridge, poker,

toe scrabble, nuclear war
 Contd..
 Perfect information
 A game in which agents can look into the complete board.
 Agents have all the information about the game, and they
can see each other moves also.
 Examples are Chess, Checkers, Go, etc.

 Imperfect information
 A game in which agents do not have all information about
the game and not aware with what's going on,
 Examples are tic-tac-toe, Battleship, blind, Bridge, etc.
 Contd..
 Deterministic games
 Those games which follow a strict pattern and set of rules for the
games, and there is no randomness associated with them.
 Examples are chess, Checkers, Go, tic-tac-toe, etc.

 Non-deterministic games
 Those games which have various unpredictable events and has a
factor of chance or luck.
 This factor of chance or luck is introduced by either dice or cards.
These are random, and each action response is not fixed.
 Also called as stochastic games.
 Example: Backgammon, Monopoly, Poker, etc.
 Zero-Sum Game
 Adversarial search which involves pure competition.
 Each agent's gain or loss of utility is exactly balanced
by the losses or gains of utility of another agent.
 One player of the game try to maximize one single
value, while other player tries to minimize it.
 Each move by one player in the game is called as ply.
 Example:
 Chess and tic-tac-toe.
 Formalization of the problem
 A game can be defined as a type of search in AI which
can be formalized of the following elements:
 Initial state
 It specifies how the game is set up at the start.
 Player(s)
 It specifies which player has moved in the state space.
 Action(s)
 It returns the set of legal moves in state space.
 Result(s, a)
 It is the transition model, which specifies the result of
moves in the state space.
 Contd..
 Terminal-Test(s)
 The state where the game ends is called terminal states.
 Terminal test is true if the game is over, else it is false at
any case.
 Utility(s, p)
 A utility function gives the final numeric value for a game
that ends in terminal states s for player p. It is also called
payoff function. (Reward after winning the game)
 For Chess, the outcomes are a win, loss, or draw and its
payoff values are +1, 0, ½. And for tic-tac-toe, utility values
are +1, -1, and 0
 Game tree
 A game tree is a tree where nodes of the tree are the
game states and Edges of the tree are the moves by
players.
Tic-Tac-Toe game tree
 Mini-Max Algorithm
 A recursive or backtracking algorithm
 The algorithm proceeds all the way down to the terminal node of the
tree, compare the values and backtrack the tree to root node to decide
the move.
 Performs a depth-first search algorithm for the exploration of
the complete game tree.
 Why not BFS??
 Best Move strategy used
 Both players will adopt best move to not allow the opponent to win
 Computes the minimax decision for the current state
 Max will try to maximize its utility (Best move)
 Min will try to minimize the utility (Worst move)
 Working of Min-Max Algorithm
 In the first step, the algorithm generates the entire game-tree
and applies the utility function to get the utility values for the
terminal states
 Contd..
 Step 1
 let's take A is the initial state of the tree. Suppose maximizer
takes first turn (normally max takes first move)
 Then min will take second move and then again max, so on

 Step2
 Start from terminal values, and find utilities for the
maximizer by comparing the values of terminal nodes
For node D max(-1,4)= 4
For Node E max(2, 6)= 6
For Node F max(-3,-5) = -3
For node G max(0, 7) = 7
Contd..
 Contd..
 Step 3
 In the next step, it's a turn for minimizer
 it will compare all nodes value, and will find the min
values.
For node B= min(4,6) = 4 For node C= min (-3, 7) =
-3
 Contd..
 Step 4
 Now it's a turn for Maximizer, and it will again choose the
maximum of all nodes value and find the maximum value for
the root node. i.e. For node A max(4, -3)= 4
 Contd..
 So, if max follows A->B->D->I, it will definitely win
with the utility of 4.
 Task
A

B C

D E F G

H I J K L M N O
-1 8 -3 -1 2 1 -3 4
 Performance Measurement
 Complete ?
 YES, It will definitely find a solution (if exist), in the finite
search tree.
 Optimal ?
 YES, if both opponents are playing optimally
 Time ?
 The worst case time complexity is O(bd). Simply O(n)
 Space ?
 The worst case space complexity is O(bd).
 Limitation of the minimax Algorithm
 The main drawback of the minimax algorithm is that
it gets really slow for complex games such as Chess,
go, etc.
 This type of games has a huge branching factor, and
the player has lots of choices to decide
 Traversing becomes complex

 Solution
 alpha-beta pruning to reduce the tree and
increase efficiency
 Alpha-Beta Pruning
 An optimization technique/ modified version of the
minimax algorithm.
 We cannot eliminate the exponential expansion of
minimax, but we can cut it to half using pruning
technique
 A technique by which without checking each node of the
game tree we can compute the correct minimax decision
 Pruning makes the algorithm fast, as sometimes it not only
prunes the leaves but also entire sub-tree that is not really
affecting the final decision but making algorithm slow
 Contd..
 Involves two threshold parameter Alpha and beta for
future expansion, so name alpha-beta pruning/Algo.
 The two-parameter can be defined as:
 Alpha: The best (highest-value) choice we have found so
far at any point along the path of Maximizer. The initial
value of alpha is -∞.
 Beta: The best (lowest-value) choice we have found so far
at any point along the path of Minimizer. The initial value
of beta is +∞.
 Cut off search by exploring less no of nodes
 If a best path is found, remaining paths are not explored
instead we cut off them
 Example
 Step 1: At the first step the, Max player will start first move
from node A (Max), will move downwards to B->D and
terminal node and find α and β values using utility values.
 Contd..
 Step 2: at node D, left value of 2 will be assigned to α. Now
it’s mean that value of D will be ≥ 2 as its max turn. Now
check other node that has value of 3 and it is turn of Max, so α
value be updated with 3 and D node will have value of 3.
 Step 3: now at node B, assign value of node D to β at node B.
as its min turn its mean node B will have value ≤ 3.
 After that, next successor E will be traversed which will get
value of α=5 from left leaf of E. as it’s max turn, its mean that
node E will have value of ≥ 5, where as node B already have a
minimum value of 3. So, there is no need to traverse the other
successor of E and hence pruned.
Contd..
 Contd..
 Step 4: now at node A, α will get value of 3. as its turn of
max, so value of A will be ≥ 3.
 Here any other path that will give value less than 3 will be pruned.
 Contd..
 Step 5:Now, node C from right side will be traversed, which
will go downward to F and F will get value of 0 for α. As it s
turn of Max so value for F will be ≥ 0. right successor of F has
value 1, so α value be updated with 1 and F will get max value
of 1.
 Step 6: F will return value of 1 to β at node C. as its turn of
Min and value of C will be ≤ 1, whereas node A already has
max value of 3, so another value generated after traversing
nodes further will return value ≤ 1. Therefore, no need to
explore further and right complete branch of C will be pruned.

 A will retain utility value of 3 i.e. optimal value for maximizer.

 Contd..
 So, So, if max follows A->B->D->I, it will definitely
win with the utility of 3.
 Complexity
 Complexity is dependent on traversal ordering of nodes. It can
be of two types:
 Worst ordering
 In some cases, alpha-beta pruning algorithm does not prune any of the
leaves of the tree, and works exactly as minimax algorithm.
 In this case, the best move occurs on the right side of the tree and it also
consumes more time because of alpha-beta factors.
 The time complexity for such an order is O(bd).
 Ideal ordering:
 The ideal ordering for alpha-beta pruning occurs when lots of pruning
happens in the tree, and best moves occur at the left side of the tree.
 We apply DFS hence, it first search left of the tree and go deep twice as
minimax algorithm in the same amount of time.
 Complexity in ideal ordering is O(bd/2).
 Task

Find the best path and optimal value for Max using alpha-beta pruning .
END