Super Tic-Tac-Toe
The game is almost the same as tic-tac-toe, but you must get 4 in a row, or in a column,
or 5 across the diagonal to win.
The board is of the shape of a cross, comprising of 5 squares, with each square of 4 x 4
size.
Player one and player two take turns to choose an empty square to place noughts and
crosses respectively.
After a player chooses an empty square, there is only ½ chance that his nought or cross
is placed at the chosen square. If the player’s choice is not accepted, the player’s move
is selected randomly with probability 1/16 by the computer from the 8 random squares
adjacent to the chosen one, with the boundaries ignored. If the random choice is
occupied or outside of the board, the player’s move is forfeited. For example, if the
chosen square is at the corner, with probability 5/16 the randomly selected square is
outside of the board.
Train an RL agent to play this game.
Requirements: Implement this using TF Agent of TensorFlow
Please make sure you 1) write tests 2) write comments 3) report that explain your
results