5.module3 ADVERSARIAL SEARCH 4

Adversarial Search
1
Outline
• Introduction
• Optimal decisions
• Minimax procedure
• α-β pruning
• Imperfect, real-time decisions
2
Introduction
• Game playing is one of the oldest sub-filed of
AI.
• Game playing involves abstract and pure form
of competition that seems to require
intelligence
• It is easy to represent the state and actions
• To implement game playing is required very
little world knowledge
3
• The most common used AI technique in game is
search
• Game playing research has contributed ideas on how
to make the best use of time to reach good decisions
• Game playing is a search problem defined by:
» Initial state of the game
» Operators defining legal moves
» Successor function
» Terminal test defining end of game states
» Goal test
» Path cost/ utility/ payoff function
4
Characteristics of game playing
• There is always an “Unpredictable” opponent:

– Due to their uncertainty
– He/she also wants to win
– Solution for each problem is a strategy, which specifies a move for
every possible opponent reply
• Time Limits:
– Games are often played under strict time constraints and therefore
must be very effectively handled
– There are special games where the two players have exactly opposite
goals.
– It can be divided into two types
» Perfect Information games (where both players have access to the same
information)
» Imperfect Information games (different information can be accessed)
5
Optimal Decisions in Games
• MIN and MAX are two game players
• First move performed by MAX then turn is given to MIN and
so on until the game is over
• A loser need to pay penalties and game points a credits are
gifted to the winning player, at end of the game
• A game as search problem and broadly defined by following
components
1. The Initial State: It is start state position on the board, which leads to
further move
2. A successor function: It is collection of (move, state) pairs, each
specifies a legal move and output state. Here actions are considered
3. Terminal Test: It is final state which declares that the game is over or
ended or terminal states
6
• Game Tree: Game tree is graphical representation of the
initial state and legal moves for each side (two players – one
by one) for a specific game
• MAX has nine possible move from initial state
• MAX with X and MIN with O symbol. The leaf nodes are
entitled by utility value
• The MAX is assumed with high values and MIN is with low
values. The best move is always determined by MAX by using
search tree
• Optimal Strategies
– A sequence of moves which achieves Win state/ Terminal state/ Goal
state in search problem
• Optimal Strategy in Game
– It is techniques which always leads to superior that any other strategy
as opponent is playing in perfect manner 7
Game tree (2-player, deterministic, turns)
8
MINIMAX Procedure
• Starting from the leaves of the tree (with final scores
with respect to one player, MAX), and go backwards
towards the root
• At each step, one player (MAX) takes the action that

leads to the highest score, while the other player
(MIN) takes the action that leads to the lowest score
• All nodes in the tree will be scored, and the path from
root to the actual result is the one on which all nodes
have the same score
9
Minimax algorithm
10
Minimax
• Perfect play for deterministic games
• Idea: choose move to position with highest minimax value
= best achievable payoff against best play
• E.g., 2-ply game:
11
Properties of minimax
• Complete? Yes (if tree is finite)
• Optimal? Yes (against an optimal opponent)
• Time complexity? O(bm)
• Space complexity? O(bm) (depth-first exploration)
12
Alpha (α)- beta (β) pruning
• The basic idea of alpha-beta cutoffs is “It is possible
to compute the correct minimax decision without
looking at every node in the search tree”
• Allow the search process to ignore portion of thee
search tree that make no difference to the final choice
• General principle of α-β pruning is
– Consider a node n somewhere in thee tree, such that a
player has a chance to move to this node
– If player has a better chance m either at the parent node of
n then n will never be reached in actual play
13
• Alpha: The highest value that the maximizer
can guarantee himself by making some move
at the current node OR at some node earlier on
the path
• Beta: The lowest value that the minimizer can
guarantee by making some move at the current
node OR at some node earlier on the path to
this node
14
α-β pruning example
15
16
17
18
19
Properties of α-β
• Pruning does not affect final result
• Good move ordering improves effectiveness of pruning
• With "perfect ordering," time complexity = O(bm/2)

 doubles depth of search
• A simple example of the value of reasoning about which

computations are relevant (a form of metareasoning)
20
Why is it called α-β?
• α is the value of the best
(i.e., highest-value) choice
found so far at any choice
point along the path for max
• If v is worse than α, max

will avoid it
 prune that branch
• Define β similarly for min
21
The α-β algorithm
22
The α-β algorithm
23

5.module3 ADVERSARIAL SEARCH 4

Uploaded by

Copyright:

Available Formats

5.module3 ADVERSARIAL SEARCH 4

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

5.module3 ADVERSARIAL SEARCH 4

Uploaded by

Copyright:

Available Formats

Adversarial Search

• There is always an “Unpredictable” opponent:

• At each step, one player (MAX) takes the action that

• Complete? Yes (if tree is finite)

• Optimal? Yes (against an optimal opponent)

• Time complexity? O(bm)

• Space complexity? O(bm) (depth-first exploration)

• Pruning does not affect final result

• Good move ordering improves effectiveness of pruning

• With "perfect ordering," time complexity = O(bm/2)

• A simple example of the value of reasoning about which

• If v is worse than α, max

 prune that branch

• Define β similarly for min

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.