Information Technology Reference
In-Depth Information
Players PX and PO make alternating moves during which each player marks the
board with a 'X' or an 'O'. Player PX uses always a 'X”, while player PO uses always
an 'O'. If any of the two players manages to mark the board with three 'X' or three
'O' to form a straight line (horizontal, or vertical, or diagonal), then that player wins.
If the board has been completely marked and none of the two players achieves a 3-
same-symbol straight line, then the game is a tie, i.e., none of the two players wins.
Figure 9 shows two winning and one tie board configurations.
O
X
O
O
X
X
O
O
X
X
X
O
X
O
O
X
X
X
O
(a) PX wins
(b) PO wins
(a) Tie
Fig. 9. Two winning and one tie Tic-Tac-To board configurations
Next we describe how the AKL can be utilized to facilitate computer learning for
the game of Tic-Tac-To, when one of the players is a computer, C, and the other
player is a human, H. In our setting, C marks the board with the symbol 'C' and H
marks the board with the symbol 'H'. If either C or H manage to achieve a straight
line of 3-same symbols, (C, C, C or H, H, H) then the corresponding player wins; if
the board has been completely marked and none of the two players achieved a straight
line with its symbols, the game is a tie and a new game may begin. In our setting, we
form a K-line for each game. The assumption is that many games are played, i.e.,
many K-lines are formed. At the beginning, before any game is played, we assume
that the computer knows what are the rules of the game (i.e., it knows that a legal
move is to mark any one empty slot of the board with its symbol 'C') but it does not
possess any strategic knowledge regarding what constitutes a good move. On the
other hand, it is assumed that H is an expert in this game. Note, for the simple game
of Tic-Tac-To, any adult human of average intelligence can be considered an expert.
The idea is that as each player (either C, or H) makes a move a K-node is formed. The
K-nodes that are formed by successive moves are connected to form a K-line. The K-
line ends with the win of one of the players, or when a tie occurs. Then a new game
begins and a new K-line is formed, and so on. Prior to the first move of the very first
game, the computer has no intelligence. Therefore, when it is its turn to move, it just
makes a random (but legal) move. But as the game progresses, and especially as more
and more games are played, the computer becomes capable of making more intelli-
gent moves by drawing on the knowledge that it has been stored in the AKL that has
been formed up to that point. The details of the K-node formation and the AKL for-
mation are described next.
Search WWH ::




Custom Search