Exemplo n.º 1
0
def sarsa(lambda_value, iteration_num):
    """

    :rtype : MSE (float)
    """
    print 'lambda:', lambda_value

    # define functions (dictionaries)
    action_value_function = ActionValue(float)
    number_state = defaultdict(int)
    number_state_action = defaultdict(int)

    # define parameters
    batch = 100
    num_zero = 10

    if lambda_value == 0. or lambda_value == 1.:
        learning_curve = []

    # iterate over iteration_num
    for episode in xrange(iteration_num):
        if episode % batch == 0:
            print '\repisode:', episode,

            if lambda_value == 0. or lambda_value == 1.:
                learning_curve.append(calculate_mse(action_value_function))

        # initialize state, action, epsilon, and eligibility-trace
        state = environment.State()
        current_dealer = state.dealer
        current_player = state.player
        epsilon = float(num_zero) / (num_zero + number_state[(current_dealer, current_player)])
        current_action = epsilon_greedy(action_value_function, state, epsilon)
        eligibility = defaultdict(int)

        while state.terminal is False:
            # count state and state_action number
            number_state[(current_dealer, current_player)] += 1
            number_state_action[(current_dealer, current_player, current_action)] += 1

            # take an action
            reward = step(state, current_action)
            if reward is None:
                # assign 0 if the match hasn't finished yet
                reward = 0
            new_dealer = state.dealer
            new_player = state.player

            # update epsilon
            epsilon = float(num_zero) / (num_zero + number_state[(new_dealer, new_player)])

            # decide an action
            new_action = epsilon_greedy(action_value_function, state, epsilon)

            # update alpha, delta, and eligibility-trace
            alpha = 1. / number_state_action[(current_dealer, current_player, current_action)]
            delta = reward + action_value_function[(new_dealer, new_player, new_action)] \
                    - action_value_function[(current_dealer, current_player, current_action)]
            eligibility[(current_dealer, current_player, current_action)] += 1

            # iterate over every state and action
            # To Be Fixed: this implementation is probably slower than vectorized way
            for key in action_value_function.keys():
                dealer, player, action = key

                # update the action value function
                action_value_function[(dealer, player, action)] \
                    += alpha * delta * eligibility[(dealer, player, action)]

                # update eligibility-trace
                eligibility[(dealer, player, action)] *= lambda_value

            # update state and action
            current_dealer = new_dealer
            current_player = new_player
            current_action = new_action

    print '\repisode:', episode
    print 'done!'

    if lambda_value == 0. or lambda_value == 1.:
        learning_curve.append(calculate_mse(action_value_function))

    # plot learning curve
    if lambda_value == 0. or lambda_value == 1.:
        x = range(0, iteration_num + 1, batch)
        pylab.title('Learning curve of Mean-Squared Error against episode number: lambda = ' + str(lambda_value))
        pylab.xlabel("episode number")
        pylab.xlim([0, iteration_num])
        pylab.xticks(range(0, iteration_num + 1, batch))
        pylab.ylabel("Mean-Squared Error")
        pylab.plot(x, learning_curve)
        pylab.show()

    # calculate MSE
    print 'calculate the Mean-Squared Error...'
    MSE = calculate_mse(action_value_function)

    ## value function
    #value_function = action_value_function.to_value_function()

    ## plot the optimal value function
    #plot_value_function(value_function, "Optimal Value Function (Sarsa)")

    return MSE
def sarsa(lambda_value, iteration_num):
    """

    :rtype : MSE (float)
    """
    print 'lambda:', lambda_value

    # define functions (dictionaries)
    action_value_function = ActionValueLinearApproximation(float)
    linear_function = LinearFunction()
    parameters_hit = np.array([0 for i in range(3 * 6)])
    parameters_stick = np.array([0 for i in range(3 * 6)])

    # define parameters
    batch = 100
    num_zero = 10
    epsilon = 0.1
    alpha = 0.01
    HIT = 0
    STICK = 1

    if lambda_value == 0. or lambda_value == 1.:
        learning_curve = []

    # iterate over iteration_num
    for episode in xrange(iteration_num):
        if episode % batch == 0:
            print '\repisode:', episode,

            if lambda_value == 0. or lambda_value == 1.:
                learning_curve.append(calculate_mse(action_value_function))

        # initialize state, action, and eligibility-trace
        state = environment.State()
        linear_function.update(state)
        current_features = linear_function.get_features()
        action = epsilon_greedy_linear(action_value_function, current_features,
                                       epsilon)
        eligibility_hit = np.array([0 for i in range(3 * 6)])
        eligibility_stick = np.array([0 for i in range(3 * 6)])

        while state.terminal is False:
            # update delta, and eligibility-trace
            if action == HIT:
                eligibility_hit = np.add(eligibility_hit,
                                         np.array(current_features))
            else:
                eligibility_stick = np.add(eligibility_stick,
                                           np.array(current_features))

            # take an action
            reward = step(state, action)
            if reward is None:
                # assign 0 if the match hasn't finished yet
                reward = 0
            linear_function.update(state)
            new_features = linear_function.get_features()

            # update delta
            delta_hit = reward - np.array(new_features).dot(parameters_hit)
            delta_stick = reward - np.array(new_features).dot(parameters_stick)

            # update Action Value Function
            if action == HIT:
                action_value_function.update_value((new_features, action),
                                                   parameters_hit)
            else:
                action_value_function.update_value((new_features, action),
                                                   parameters_stick)

            # update delta, parameters, and eligibility-trace
            if action == HIT:
                delta_hit += action_value_function[(new_features, HIT)]
            else:
                delta_stick += action_value_function[(new_features, STICK)]
            parameters_hit = np.add(parameters_hit,
                                    alpha * delta_hit * eligibility_hit)
            parameters_stick = np.add(parameters_stick,
                                      alpha * delta_stick * eligibility_stick)
            eligibility_hit = eligibility_hit * lambda_value
            eligibility_stick = eligibility_stick * lambda_value

            # decide an action
            action = epsilon_greedy_linear(action_value_function, new_features,
                                           epsilon)

            # update state and action
            current_features = new_features

    print '\repisode:', episode
    print 'done!'

    if lambda_value == 0. or lambda_value == 1.:
        learning_curve.append(calculate_mse(action_value_function))

    # plot learning curve
    if lambda_value == 0. or lambda_value == 1.:
        x = range(0, iteration_num + 1, batch)
        pylab.title(
            'Learning curve of Mean-Squared Error against episode number: lambda = '
            + str(lambda_value))
        pylab.xlabel("episode number")
        pylab.xlim([0, iteration_num])
        pylab.xticks(range(0, iteration_num + 1, batch))
        pylab.ylabel("Mean-Squared Error")
        pylab.plot(x, learning_curve)
        pylab.show()

    # calculate MSE
    print 'calculate the Mean-Squared Error...'
    MSE = calculate_mse(action_value_function)

    ## value function
    #value_function = action_value_function.to_value_function()

    ## plot the optimal value function
    #plot_linear_value_function(action_value_function, "Optimal Value Function (Linear Approximation)")

    return MSE
def sarsa(lambda_value, iteration_num):
    """

    :rtype : MSE (float)
    """
    print 'lambda:', lambda_value

    # define functions (dictionaries)
    action_value_function = ActionValueLinearApproximation(float)
    linear_function = LinearFunction()
    parameters_hit = np.array([0 for i in range(3 * 6)])
    parameters_stick = np.array([0 for i in range(3 * 6)])

    # define parameters
    batch = 100
    num_zero = 10
    epsilon = 0.1
    alpha = 0.01
    HIT = 0
    STICK = 1

    if lambda_value == 0. or lambda_value == 1.:
        learning_curve = []

    # iterate over iteration_num
    for episode in xrange(iteration_num):
        if episode % batch == 0:
            print '\repisode:', episode,

            if lambda_value == 0. or lambda_value == 1.:
                learning_curve.append(calculate_mse(action_value_function))

        # initialize state, action, and eligibility-trace
        state = environment.State()
        linear_function.update(state)
        current_features = linear_function.get_features()
        action = epsilon_greedy_linear(action_value_function, current_features, epsilon)
        eligibility_hit = np.array([0 for i in range(3 * 6)])
        eligibility_stick = np.array([0 for i in range(3 * 6)])

        while state.terminal is False:
            # update delta, and eligibility-trace
            if action == HIT:
                eligibility_hit = np.add(eligibility_hit, np.array(current_features))
            else:
                eligibility_stick = np.add(eligibility_stick, np.array(current_features))

            # take an action
            reward = step(state, action)
            if reward is None:
                # assign 0 if the match hasn't finished yet
                reward = 0
            linear_function.update(state)
            new_features = linear_function.get_features()

            # update delta
            delta_hit = reward - np.array(new_features).dot(parameters_hit)
            delta_stick = reward - np.array(new_features).dot(parameters_stick)

            # update Action Value Function
            if action == HIT:
                action_value_function.update_value((new_features, action), parameters_hit)
            else:
                action_value_function.update_value((new_features, action), parameters_stick)

            # update delta, parameters, and eligibility-trace
            if action == HIT:
                delta_hit += action_value_function[(new_features, HIT)]
            else:
                delta_stick += action_value_function[(new_features, STICK)]
            parameters_hit = np.add(parameters_hit, alpha * delta_hit * eligibility_hit)
            parameters_stick = np.add(parameters_stick, alpha * delta_stick * eligibility_stick)
            eligibility_hit = eligibility_hit * lambda_value
            eligibility_stick = eligibility_stick * lambda_value

            # decide an action
            action = epsilon_greedy_linear(action_value_function, new_features, epsilon)

            # update state and action
            current_features = new_features

    print '\repisode:', episode
    print 'done!'

    if lambda_value == 0. or lambda_value == 1.:
        learning_curve.append(calculate_mse(action_value_function))

    # plot learning curve
    if lambda_value == 0. or lambda_value == 1.:
        x = range(0, iteration_num + 1, batch)
        pylab.title('Learning curve of Mean-Squared Error against episode number: lambda = ' + str(lambda_value))
        pylab.xlabel("episode number")
        pylab.xlim([0, iteration_num])
        pylab.xticks(range(0, iteration_num + 1, batch))
        pylab.ylabel("Mean-Squared Error")
        pylab.plot(x, learning_curve)
        pylab.show()

    # calculate MSE
    print 'calculate the Mean-Squared Error...'
    MSE = calculate_mse(action_value_function)

    ## value function
    #value_function = action_value_function.to_value_function()

    ## plot the optimal value function
    #plot_linear_value_function(action_value_function, "Optimal Value Function (Linear Approximation)")

    return MSE
Exemplo n.º 4
0
            dealer = state.dealer
            player = state.player

            # update epsilon
            epsilon = float(num_zero) / (num_zero +
                                         number_state[(dealer, player)])

            # decide an action
            action = epsilon_greedy(action_value_function, state, epsilon)

            # count state and state_action number
            number_state[(dealer, player)] += 1
            number_state_action[(dealer, player, action)] += 1

            # take an action
            reward = step(state, action)
            if reward is None:
                # assign 0 if the match hasn't finished yet
                reward = 0

            # update the action value function
            action_value_function[(dealer, player, action)] += \
                (1. / number_state_action[(dealer, player, action)]) \
                * (reward - action_value_function[(dealer, player, action)])

    print '\repisode:', episode
    print 'done!'

    # calculate value function
    value_function = action_value_function.to_value_function()
Exemplo n.º 5
0
        while state.terminal is False:
            dealer = state.dealer
            player = state.player

            # update epsilon
            epsilon = float(num_zero) / (num_zero + number_state[(dealer, player)])

            # decide an action
            action = epsilon_greedy(action_value_function, state, epsilon)

            # count state and state_action number
            number_state[(dealer, player)] += 1
            number_state_action[(dealer, player, action)] += 1

            # take an action
            reward = step(state, action)
            if reward is None:
                # assign 0 if the match hasn't finished yet
                reward = 0

            # update the action value function
            action_value_function[(dealer, player, action)] += \
                (1. / number_state_action[(dealer, player, action)]) \
                * (reward - action_value_function[(dealer, player, action)])

    print '\repisode:', episode
    print 'done!'

    # calculate value function
    value_function = action_value_function.to_value_function()