COS3751
Assignment
2 2023 -
DUE 21
July 2023
QUESTIONS AND ANSWERS
, QUESTION 1
What is utility function and why is it used in adversarial searches.
Also called an objective function or a payoff function.
Defines the final numeric value for a game that ends in a terminal state s for a player p.
An agent’s utility function is essentially an internalisation of the performance measure.
It captures the preferences of an agent and assigns a single number to express the
desirability of a state.
The utility function allows us to determine the apparent desirability of a state
Explain how forward pruning works. Provide at least one approach to
forward pruning in your explanation, as well as a problem that may be
encountered with forward pruning.
Allows us to ignore portions of the search tree that make no difference to the final choice.
Forward pruning, meaning that some moves at a given node are pruned immediately
without consideration.
Beam search: on each ply, consider only a “beam” of the n best moves rather than
considering all possible moves. This approach is rather dangerous because there’s no
guarantee that the best move won’t be pruned away.
Does the order in which nodes are examined in minimax matter? Explain your
answer.
Yes.
Minimax algorithm is a depth first, so at any one time we just have to consider the nodes
along a single path in the tree.
Also it uses a simple recursive computation of the minimax values of each successor state,
directly implementing the defining equations.
The right order of nodes could result in pruning earlier. For example: if a _value of 2 (low
values are worse for MAX) is returned for the first child of a MIN node, and a subsequent
child’s child returns 5, then no more children of the child need to be examined. On the other
hand, if that first child returned 10, and the following child returned 5, all the child’s children
may have to be examined.
Assignment
2 2023 -
DUE 21
July 2023
QUESTIONS AND ANSWERS
, QUESTION 1
What is utility function and why is it used in adversarial searches.
Also called an objective function or a payoff function.
Defines the final numeric value for a game that ends in a terminal state s for a player p.
An agent’s utility function is essentially an internalisation of the performance measure.
It captures the preferences of an agent and assigns a single number to express the
desirability of a state.
The utility function allows us to determine the apparent desirability of a state
Explain how forward pruning works. Provide at least one approach to
forward pruning in your explanation, as well as a problem that may be
encountered with forward pruning.
Allows us to ignore portions of the search tree that make no difference to the final choice.
Forward pruning, meaning that some moves at a given node are pruned immediately
without consideration.
Beam search: on each ply, consider only a “beam” of the n best moves rather than
considering all possible moves. This approach is rather dangerous because there’s no
guarantee that the best move won’t be pruned away.
Does the order in which nodes are examined in minimax matter? Explain your
answer.
Yes.
Minimax algorithm is a depth first, so at any one time we just have to consider the nodes
along a single path in the tree.
Also it uses a simple recursive computation of the minimax values of each successor state,
directly implementing the defining equations.
The right order of nodes could result in pruning earlier. For example: if a _value of 2 (low
values are worse for MAX) is returned for the first child of a MIN node, and a subsequent
child’s child returns 5, then no more children of the child need to be examined. On the other
hand, if that first child returned 10, and the following child returned 5, all the child’s children
may have to be examined.
