Exemplo n.º 1
0
    def depth_3_cnn(self, X_train, Y_train, X_test, Y_test, num_classes,
                    nb_epoch, verbose, validation_split, batch_size, filterNum,
                    dim1, dim2, img_row, img_col, img_channel):
        model = Sequential()
        model.add(
            Conv2D(filterNum, (dim1, dim2),
                   padding='same',
                   input_shape=(img_row, img_col, img_channel),
                   activation='relu'))
        model.add(MaxPooling2D(poo_size=(4, 4)))
        model.add(
            Conv2D((filterNum * 2), (dim1, dim2),
                   padding='same',
                   activation='relu'))
        model.add(MaxPooling2D(pool_size=(2, 2)))
        model.add(
            Conv2D((filterNum * 4), (dim1, dim2),
                   padding='same',
                   activation='relu'))
        model.add(Flatten())
        model.add(Dense(512), activation='relu')
        model.add(Dense(num_classes), activation='softmax')
        model.model(loss='categorical_crossentropy',
                    optimizer=RMSProp(),
                    metric=['accuracy'])
        model.summary()

        model.fit(X_train,
                  Y_train,
                  batch_size=batch_size,
                  epochs=nb_epoch,
                  validation_split=validation_split,
                  verbose=verbose)
        score = model.evaluate(X_test,
                               Y_test,
                               batch_size=batch_size,
                               verbose=verbose)

        #Save Model Json
        model_json = model.to_json()
        with open("CNN01_model.json", "w") as json_file:
            json_file.write(model_json)
        #Save Model H5
        model.save_weights("CNN01_model.h5")
        accuracy = score[1]
        accuracy = accuracy * 100

        return accuracy
Exemplo n.º 2
0

Sequential.model = _model_evaluation
rho_regressor = Sequential()
rho_regressor.add(Dense(4, input_dim=n_q_regressors_weights, init='uniform',
                        activation=ACTIVATION))
rho_regressor.add(
    Dense(n_q_regressors_weights, init='uniform', activation='linear'))
rho_regressor.compile(loss='mse', optimizer='rmsprop')

import theano
import theano.tensor as T

theta = T.matrix()

res = rho_regressor.model(theta)


# rho_regressor.fit(None, None)
##########################################

def terminal_evaluation(old_theta, new_theta, tol_theta=1e-2):
    if increment_base_termination(old_theta, new_theta, 2, tol_theta):
        estimator = LQG_Q()
        estimator.omega = new_theta[0]
        agent = Algorithm(estimator, state_dim, action_dim,
                          discrete_actions, mdp.gamma, mdp.horizon)
        agent._iteration = 1
        initial_states = np.array([[1, 2, 5, 7, 10]]).T
        values = evaluation.evaluate_policy(mdp, agent,
                                            initial_states=initial_states)
Exemplo n.º 3
0
rho_regressor = Sequential()
rho_regressor.add(
    Dense(4,
          input_dim=n_q_regressors_weights,
          init='uniform',
          activation=ACTIVATION))
rho_regressor.add(
    Dense(n_q_regressors_weights, init='uniform', activation='linear'))
rho_regressor.compile(loss='mse', optimizer='rmsprop')

import theano
import theano.tensor as T

theta = T.matrix()

res = rho_regressor.model(theta)

# rho_regressor.fit(None, None)
##########################################


def terminal_evaluation(old_theta, new_theta, tol_theta=1e-2):
    if increment_base_termination(old_theta, new_theta, 2, tol_theta):
        estimator = LQG_Q()
        estimator.omega = new_theta[0]
        agent = Algorithm(estimator, state_dim, action_dim, discrete_actions,
                          mdp.gamma, mdp.horizon)
        agent._iteration = 1
        initial_states = np.array([[1, 2, 5, 7, 10]]).T
        values = evaluation.evaluate_policy(mdp,
                                            agent,