TensorFlow2.0 模型构建

太过爱你忘了你带给我的痛 2022-10-20 01:29 149阅读 0赞

##### 模型构建 #####

# -*- coding: utf-8 -*-
    # @Time : 2021/7/12 8:36
    # @Author : 
    import matplotlib.pyplot as plt
    import pandas as pd
    import tensorflow as tf
    
    fashion_mnist = tf.keras.datasets.fashion_mnist
    (x_train_all, y_train_all), (x_test, y_test) = fashion_mnist.load_data()
    
    # 训练集拆分为训练集和验证集
    x_valid, x_train = x_train_all[:5000], x_train_all[5000:]
    y_valid, y_train = y_train_all[:5000], y_train_all[5000:]
    print(x_valid.shape, y_valid.shape)
    print(x_train.shape, y_train.shape)
    print(x_test.shape, y_test.shape)
    
    #### 模型构建
    # Sequential把一系列层次堆叠
    # tf.keras.models.Sequential()
    
    """ model = tf.keras.Sequential() # 输入层 # 将输入图片进行展开,输入是28*28的图像 # Flatten展平 将28*28二维矩阵展平为28*28一维向量 model.add(tf.keras.layers.Flatten(input_shape=[28, 28])) # 全连接层(神经网络里最普通的一种神经网络,有层次,下一层所有单元和上一层单元都进行一一连接) # 单元数300 activation激活函数 model.add(tf.keras.layers.Dense(300, activation="relu")) model.add(tf.keras.layers.Dense(100, activation="relu")) model.add(tf.keras.layers.Dense(10, activation="softmax")) """
    # relu: y=max(0,x)
    # softmax: 将向量变成概率分布. x = [x1, x2, x3],
    # y=[e^x1/sum, e^x2/sum, e^s3/sum],sum=e^x1 + e^x2 + e^x3
    
    model = tf.keras.Sequential([
        tf.keras.layers.Flatten(input_shape=[28, 28]),
        tf.keras.layers.Dense(300, activation="relu"),
        tf.keras.layers.Dense(100, activation="relu"),
        tf.keras.layers.Dense(10, activation="softmax")
    ])
    # loss损失函数
    # reason for sparse: y->index. y->one_hot->[]
    # 如果y是数 sparse_categorical_crossentropy
    # 如果y是向量categorical_crossentropy
    # optimizer模型的求解方法
    # metrics 指标
    model.compile(loss="sparse_categorical_crossentropy",
                  # optimizer="sgd",
                  optimizer="adam",
                  metrics=["accuracy"])
    
    # 查看模型有多少层
    print(model.layers)
    
    # 查看模型概况
    print(model.summary())
    # 第一层[None,784]:样本数*784的矩阵
    # 经过全连接层之后变成[None, 300]:样本数*300的矩阵
    # W(矩阵) b(偏置)
    # 784*300+300=235500
    # [None,784] * W + b -> [None, 300] W.shape [784, 300], b = [300]
    
    # 训练函数
    # epochs遍历训练集次数
    history = model.fit(x_train, y_train, epochs=10,
                        validation_data=(x_valid, y_valid))
    
    type(history)
    
    # 查看相应数据
    print(history.history)
    
    
    def plot_learning_curves(history):
        pd.DataFrame(history.history).plot(figsize=(8, 5))
        plt.grid(True)
        plt.gca().set_ylim(0, 1)
        plt.show()
    
    
    plot_learning_curves(history)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0NvY2t0YWlsX3B5_size_16_color_FFFFFF_t_70]  
[2-4实战分类之模型构建][2-4]

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0NvY2t0YWlsX3B5_size_16_color_FFFFFF_t_70]: /images/20221004/3966f529fbec475c8142e326aa3ed336.png
[2-4]: https://blog.csdn.net/weixin_46815330/article/details/109842338