Пример #1
0
def main():
    batch_size = 32  # the number of training examples in one forward/backward pass
    num_classes = 10  # number of cifar-10 dataset classes
    epochs = 3  # number of forward and backward passes of all the training examples
    '''
        dataset contains the hyper parameters for loading data and the dataset:
            dataset = {
                'batch_size': batch_size,
                'num_classes': num_classes,
                'epochs': epochs,
                'x_train': x_train,
                'x_test': x_test,
                'y_train': y_train,
                'y_test': y_test
            }
    '''
    dataset = load_dataset(batch_size, num_classes, epochs)

    num_population = 4
    num_generation = 4
    num_offspring = 2

    # plot the best model obtained
    optCNN = genetic_algorithm(num_population, num_generation, num_offspring,
                               dataset)

    # plot the training and validation loss and accuracy
    num_epoch = 3
    model = optCNN.build_model()
    model.compile(optimizer='rmsprop',
                  loss='categorical_crossentropy',
                  metrics=['accuracy'])
    history = model.fit(dataset['x_train'],
                        dataset['y_train'],
                        batch_size=dataset['batch_size'],
                        epochs=num_epoch,
                        validation_data=(dataset['x_test'], dataset['y_test']),
                        shuffle=True)
    optCNN.model = model  # model
    optCNN.fitness = history.history['val_loss'][-1]  # fitness

    print("\n\n-------------------------------------")
    print("The initial CNN has been evolved successfully in the individual",
          optCNN.name)
    print("-------------------------------------\n")
    daddy = load_network('parent_0')
    model = tf.keras.models.load_model('parent_0.h5')
    print("\n\n-------------------------------------")
    print("Summary of initial CNN")
    print(model.summary())
    print("Fitness of initial CNN:", daddy.fitness)

    print("\n\n-------------------------------------")
    print("Summary of evolved individual")
    print(optCNN.model.summary())
    print("Fitness of the evolved individual:", optCNN.fitness)
    print("-------------------------------------\n")

    plot_training(history)
Пример #2
0
    def asexual_reproduction(self, it, dataset):

        # if the individual already exists, just load it
        if os.path.isfile('net_' + str(it) + '.h5'):
            print("\n-------------------------------------")
            print("Loading individual net_" + str(it))
            print("--------------------------------------\n")
            individual = load_network('net_' + str(it))
            model = tf.keras.models.load_model(individual.name + '.h5')
            print("SUMMARY OF", individual.name)
            print(model.summary())
            print("FITNESS: ", individual.fitness)
            return individual

        # otherwise, create the individual by mutating the parent
        individual = Network(it)

        print("\n-------------------------------------")
        print("\nCreating individual", individual.name)
        print("--------------------------------------\n")

        individual.block_list = deepcopy(
            self.block_list)  # copy the layer list from parent

        print("----->Strong Mutation")
        individual.block_mutation(dataset)  # mutate a block
        individual.layer_mutation(dataset)  # mutate a layer
        individual.parameters_mutation()  # mutate some parameters

        model = individual.build_model()

        if model == -1:
            return self.asexual_reproduction(it, dataset)

        individual.train_and_evaluate(model, dataset)

        return individual
Пример #3
0
    def asexual_reproduction(self, it, dataset):

        # 如果个体存在就读取它
        if os.path.isfile('net_' + str(it) + '.h5'):
            print("\n-------------------------------------")
            print("Loading individual net_" + str(it))
            print("--------------------------------------\n")
            individual = load_network('net_' + str(it))
            model = tf.keras.models.load_model(individual.name + '.h5')
            print("SUMMARY OF", individual.name)
            print(model.summary())
            print("FITNESS: ", individual.fitness)
            return individual

        # 否则通过变异父代来产生个体
        individual = Network(it)

        print("\n-------------------------------------")
        print("\nCreating individual", individual.name)
        print("--------------------------------------\n")

        individual.block_list = deepcopy(self.block_list)           

        print("----->Strong Mutation")
        individual.block_mutation(dataset)                          # 块变异
        individual.layer_mutation(dataset)                          # 层变异
        individual.parameters_mutation()                            # 参数变异

        model = individual.build_model()

        if model == -1:
            return self.asexual_reproduction(it, dataset)

        individual.train_and_evaluate(model, dataset)

        return individual
Пример #4
0
def compute_parent(dataset):
    if os.path.isfile('parent_0.h5'):
        daddy = load_network('parent_0')
        model = tf.keras.models.load_model('parent_0.h5')
        print("Loading parent_0")
        print("SUMMARY OF", daddy.name)
        print(model.summary())
        print("FITNESS:", daddy.fitness)
        return daddy

    daddy = Network(0)

    layerList1 = [
        Convolutional(filters=32,
                      filter_size=(3, 3),
                      stride_size=(1, 1),
                      padding='same',
                      input_shape=dataset['x_train'].shape[1:]),
        Convolutional(filters=32,
                      filter_size=(3, 3),
                      stride_size=(1, 1),
                      padding='valid',
                      input_shape=dataset['x_train'].shape[1:])
    ]
    layerList2 = [
        Pooling(pool_size=(2, 2), stride_size=(2, 2), padding='same')
    ]
    daddy.block_list.append(Block(0, 0, layerList1, layerList2))
    '''
    layerList1 = [
        Convolutional(filters=64, filter_size=(3, 3), stride_size=(1, 1), padding='same',
                      input_shape=dataset['x_train'].shape[1:]),
        Convolutional(filters=64, filter_size=(3, 3), stride_size=(1, 1), padding='valid',
                      input_shape=dataset['x_train'].shape[1:])
    ]
    layerList2 = [
        Pooling(pool_size=(2, 2), stride_size=(2, 2), padding='same'),
        Dropout(rate=0.25)
    ]
    daddy.block_list.append(Block(1, 1, layerList1, layerList2))
    
    layerList1 = [
        Convolutional(filters=128, filter_size=(3, 3), stride_size=(1, 1), padding='same',
                      input_shape=dataset['x_train'].shape[1:]),
        Convolutional(filters=128, filter_size=(3, 3), stride_size=(1, 1), padding='valid',
                      input_shape=dataset['x_train'].shape[1:])
    ]
    layerList2 = [
        Pooling(pool_size=(2, 2), stride_size=(2, 2), padding='same'),
        Dropout(rate=0.25)
    ]
    daddy.block_list.append(Block(1, 1, layerList1, layerList2))
    '''
    layerList1 = [
        FullyConnected(units=128, num_classes=dataset['num_classes'])
    ]
    layerList2 = []
    daddy.block_list.append(Block(2, 1, layerList1, layerList2))

    model = daddy.build_model()
    daddy.train_and_evaluate(model, dataset)
    return daddy
Пример #5
0
from model.network import network
from utilities import save_network, load_network
from networkx import to_numpy_array

net = network(10)
print('net gen')
tmp_net = to_numpy_array(net.network_plot)
print(tmp_net)

save_network(net)
print('save done')

net_cpy = load_network()
print(to_numpy_array(net_cpy))
print('load done')