2021-11-15

这是深度学习Keras框架学习的一个示例

import numpy as np
from keras.datasets import imdb
from keras import layers
from keras import models
from keras import optimizers
import matplotlib.pyplot as plt

(train_data, train_labels),(test_data, test_labels) = imdb.load_data(num_words=1000)

# 获取将将单词映射为整数的字典
word_index = imdb.get_word_index()



# One-Host向量化序列
def vectorize_sequences(sequences,dimension = 10000):
    results = np.zeros((len(sequences), dimension))
    for i, sequence in enumerate(sequences):
        results[i, sequence] = 1.
    return  results

if __name__ == '__main__':
    # 颠倒字典，以值为键，以键为值
    reverse_word_index = dict(
        [(value, key) for (key, value) in word_index.items()]
    )
    # 解码第一条数据
    decode_review = ' '.join([reverse_word_index.get(i - 3, '?') for i in train_data[0]])
    # 使用Onehost方式向量化训练数据和测试数据
    x_train = vectorize_sequences(train_data)
    x_test = vectorize_sequences(test_data)
    # 向量化标签
    y_train = np.asarray(train_labels).astype('float32')
    y_test = np.asarray(test_labels).astype('float32')

    # 构建网络
    model = models.Sequential()
    model.add(layers.Dense(16, activation='relu', input_shape=(10000,)))
    model.add(layers.Dense(16, activation='relu'))
    model.add(layers.Dense(1, activation='sigmoid'))

    # 编译模型
    model.compile(optimizer='rmsprop',loss='binary_crossentropy',metrics=['accuracy'])

    # 验证模型
    x_val = x_train[:10000]
    partial_x_train = x_train[10000:]
    y_val = y_train[:10000]
    partial_y_train = y_train[10000:]

    history = model.fit(partial_x_train,
                        partial_y_train,
                        epochs=50,
                        batch_size=512,
                        validation_data=(x_val, y_val))
    history_dict = history.history
    loss_values = history_dict['loss']
    val_loss_values = history_dict['val_loss']

    epochs = range(1, len(loss_values) + 1)

    # 绘制训练损失和验证损失
    plt.plot(epochs, loss_values, 'bo', label='Training loss')
    plt.plot(epochs, val_loss_values, 'b', label='Validation loss')
    plt.title('Training and validation loss')
    plt.xlabel('Epochs')
    plt.ylabel('Loss')
    plt.legend()
    plt.show()

实验结果：

此神经网络是书上的一个示例，我故意设置了50此迭代次数，观察随着迭代次数的增加，神经网络在训练集和测试集上的表现。由上图可见，在第10轮迭代附近时，神经网络在测试集上的效果已达最佳，迭代继续，神经网络在训练集上表现越来越好，可在测试集上的表现却越来越差，陷入了过拟合。