ttt.models package¶

Submodules¶

ttt.models.action module¶

class ttt.models.action.Action[source]¶

Bases: enum.Enum

Models all possible actions.

bottom_center = 7¶

bottom_left = 6¶

bottom_right = 8¶

middle_center = 4¶

middle_left = 3¶

middle_right = 5¶

top_center = 1¶

top_left = 0¶

top_right = 2¶

ttt.models.agent module¶

class ttt.models.agent.Agent[source]¶

Bases: object

Class modeling a player

grid¶

Representation of the board as an array of integers.

Type: numpy.ndarray

current_state¶

Current game state

Type: ttt.models.State

update_grid(grid: numpy.ndarray) → None[source]¶

Updates grid and current_state with the given value.

Parameters: grid (numpy.ndarray) – Representation of the board as an array of integers.

class ttt.models.agent.CPUAgent[source]¶

Bases: ttt.models.agent.Agent

Class modeling a CPU player.

states¶

List of states. Representation of the game as a Markov Decision Process.

Type: list

add_state(state: ttt.models.state.State) → None[source]¶

Adds the given state to the MDP.

Parameters: state (ttt.models.State) – State to be added.

deserialize(json_data: dict) → None[source]¶

Dumps specified json_data to the agent.

Parameters: json_data (dict) – Dictionary with the following format: {mdp: dict}

get_best_move() → ttt.models.action.Action[source]¶

Generates a new move using the previous knowledge.

Returns: move
Return type: ttt.models.Action

get_random_move() → ttt.models.action.Action[source]¶

Generates a random move.

Returns: move
Return type: ttt.models.Action

get_state(grid: numpy.ndarray) → ttt.models.state.State[source]¶

Gets the state with the given grid.

Parameters: grid (numpy.ndarray) – Representation of the board.
Returns: state – State containing the given grid.
Return type: ttt.models.State

has_state(state: ttt.models.state.State) → bool[source]¶

Checks if MDP has registered the given state :param state: State to be checked :type state: ttt.models.State

Returns: is_registered
Return type: bool

load(path: str) → None[source]¶

Loads a JSON file containing agent’s params.

Parameters: path (str) – Path of the JSON file.

save(path: str) → None[source]¶

Saves model’s params as a JSON file.

Parameters: path (str) – Where to save agent params.

serialize() → dict[source]¶

Serializes this Agent into a dict.

Returns: serializable_dict
Return type: dict

update_grid(grid: numpy.ndarray) → None[source]¶

Updates grid, current_state and mdp with the given value.

Parameters: grid (numpy.ndarray) – Representation of the board as an array of integers.

update_state(state: ttt.models.state.State) → None[source]¶

Updates the MDPs state with the given one.

Parameters: state (ttt.models.State) – State to update MDPs with.

class ttt.models.agent.HumanAgent[source]¶

Bases: ttt.models.agent.Agent

Class modeling a human player.

grid¶

Representation of the board as list of integers.

Type: list

get_next_move() → ttt.models.action.Action[source]¶

Gets the user move.

Returns: move
Return type: int

ttt.models.game module¶

class ttt.models.game.Game(player_1: ttt.models.agent.Agent, player_2: ttt.models.agent.Agent)[source]¶

Bases: object

Class modeling.

grid¶

Type: numpy.ndarray

player_1¶

Type: ttt.models.Agent

player_2¶

Type: ttt.models.Agent

result¶

Type: int

grid_is_full¶

Type: bool

game_sequence¶

Type: list

Parameters

player_1 (ttt.models.Agent) –
player_2 (ttt.models.Agent) –

apply_move(move: ttt.models.action.Action, player_id: int) → None[source]¶

Commits the move over the board.

Parameters

move (ttt.models.Action) – Move to be performed.
player_id (int) – Value to apply.

check_diagonal_win() → None[source]¶: Checks if there is 3 consecutives marks of the same player diagonally

check_horizontal_win() → None[source]¶: Checks if there is 3 consecutives marks of the same player horizontally.

check_if_grid_is_full() → None[source]¶: Checks if there is no more possible move.

check_vertical_win() → None[source]¶: Checks if there is 3 consecutives marks of the same player vertically.

check_victory() → None[source]¶: Checks if there is 3 consecutives marks of the same player.

print_board() → dict[source]¶: Displays the current grid.

ttt.models.reward module¶

class ttt.models.reward.Reward[source]¶

Bases: object

Models all possible rewards.

static defeat(i, lr)[source]¶

Computes the reward for a given state when the result was a defeat.

Parameters

i (int) – Position in which the state was reached
lr (float) – Learning rate

Returns

reward

Return type

float

static tie(i, lr)[source]¶

Computes the reward for a given state when the result was a tie.

Parameters

i (int) – Position in which the state was reached
lr (float) – Learning rate

Returns

reward

Return type

float

static win(i, lr)[source]¶

Computes the reward for a given state when the result was a win.

Parameters

i (int) – Position in which the state was reached
lr (float) – Learning rate

Returns

reward

Return type

float

ttt.models.state module¶

class ttt.models.state.State(grid: numpy.ndarray = None)[source]¶

Bases: object

Class modelling the states of the game

Parameters: grid (numpy.ndarray) – Representation of the board.

grid¶

Representation of the board.

Type: numpy.ndarray

next_states_values¶

The weight of the transition to a state.

Type: numpy.ndarray

next_states_transitions¶

The next move to move to the next state.

Type: numpy.ndarray

deserialize(json_data: dict) → None[source]¶

Dumps specified json_data to the state.

Parameters: json_data (dict) – Dictionary with the following format: {grid: list, next_state_values: list, next_state_transitions: list}

get_best_move() → ttt.models.action.Action[source]¶: Decides on the best available move based on the transition weights.

serialize() → dict[source]¶

Serializes this State into a dict.

Returns: serializable_dict
Return type: dict

update_transition_weight(transition: ttt.models.action.Action, increase: float) → None[source]¶

Updates the transition weights.

Parameters

transition (int) – Transition move.
increase (float) – Value to be added