• 【人工智能】神经网络八股

    使用八股搭建神经网络

    目录

    六步法搭建网络

    用tensorflow API：tf.leras搭建网络八股

    1. import # 导入相关模块
    2.  
    3. train,test # 告知要喂入网络的训练集和测试集。即指定训练集的输入特征x_train和训练集的标签y_train；指定测试集的输入特征x_test和测试集的标签y_test
    4.  
    5. model=tf.keras.models.Sequential # 在Sequential()中搭建网络结构，逐层描述每层网络，相当于走了一遍前向传播
    6.  
    7. model.compile # 在compile()中配置训练方法，告知训练时选择哪种优化器，选择哪个损失函数，选择哪种评测指标
    8.  
    9. model.fit # 在fit()中执行训练过程,告知训练集和测试集的输入标签和特征，告知，每个batch是多少，告知要迭代多少次数据集。
    10.  
    11. model.summary # 用summary()打印出网络的结构和参数统计

    (1)tf.keras.models.Sequential

    可以搭建出上层输出下层输入的神经网络结构，但是无法写出一些带有跳连的非顺序网络结构

    model = tf.keras.models.Sequential([网络结构]) # 描述各层网络

    网络结构举例：

    拉直层：tf.keras.layers.Flatten()

    这一层不含计算，只是形状转换。把输入特征拉直，变为一维数组

    全连接层：tf.keras.layers.Dense(神经元个数、activation="激活函数",kernel_regularizer=哪种正则化)

    activate(字符串给出)可选：relu,softmax,tanh,sigmoid,

    kernel_regularizer可选：,tf.keras.regularizersl1(),tf.keras.regularizers.l2()

    卷积层：tf.keras.layers.Conv2D(filters=卷积核个数,kernel_size=卷积核尺寸,strides=卷积步长,padding="valid"or"same")

    LSTM层：tf.keras.layers.LSTM()

    (2)model.compile

    model.compile(optimizer=优化器,loss=损失函数,metrics=["准确率"])

    Optimizer可选

    "sgd" or tf.keras.optimizers.SGD(lr=学习率,momentum=动量参数)

    "adagrad" or tf.keras.optimizers.Adagrad(lr=学习率)

    "adadelta" or tf.keras.optimizers.Adadelta(lr=学习率)

    "adam" or tf.keras.optimisers.Adam(lr=学习率,beta_1=0.9,beta_2=0.999)

    建议初学者使用左边这些字符串形式的优化器名字

    loss可选

    "mes" or tf.keras.losses.MeanSquaredError()

    "spaese_categorical_crossentropy" or tf.keras.losses.SparseCategoricalCrossentropy(from_logits=false)

    Metrics可选

    "accuracy":y_和y都是数值
    "categorical_accuracy":y_和y都是独热码(概率分布)
    "spare_categorical_accuracy":y_是数值，y是独热码(概率分布)

    (3)model.fit

    fit()执行训练过程

    1. model.fit(训练集的输入特征,训练集的标签,
    2.          batch_size=?,epochs=?,
    3.           validation_data=(测试集的输入特征,测试集的标签),
    4.           validation_split=从训练集划分多少比例给测试集,
    5.           validation_freq=多少次epoch测试一次
    6.          )

    batch_size:每次喂入神经网络的样本数
    epochs:要迭代多少次数据集
    validation_data 和 validation_split 二者选择其一使用
    validation_freq:没多少次epoch迭代使用测试集验证一次结果

    (4)model.summary

    summary() 可以打印网络的结构和参数统计

    总参数：Total params
    可训练参数：Trainable params
    不可训练参数：Non-Trainable params

    源码

    1. import tensorflow as tf
    2. from sklearn import datasets
    3. import numpy as np
    4.  
    5. x_train = datasets.load_iris().data
    6. y_train = datasets.load_iris().target
    7.  
    8. np.random.seed(116)
    9. np.random.shuffle(x_train)
    10. np.random.seed(116)
    11. np.random.shuffle(y_train)
    12. tf.random.set_seed(116)
    13.  
    14. model = tf.keras.models.Sequential([
    15.     tf.keras.layers.Dense(3, activation='softmax', kernel_regularizer=tf.keras.regularizers.l2())
    16. ])
    17.  
    18. model.compile(optimizer=tf.keras.optimizers.SGD(lr=0.1),
    19.               loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    20.               metrics=['sparse_categorical_accuracy'])
    21.  
    22. model.fit(x_train, y_train, batch_size=32, epochs=500, validation_split=0.2, validation_freq=20)
    23.  
    24. model.summary()
    1. Epoch 1/500
    2. 4/4 [==============================] - 0s 748us/step - loss: 2.3173 - sparse_categorical_accuracy: 0.3417
    3. 中间运行结果省略
    4. Epoch 500/500
    5. 4/4 [==============================] - 0s 3ms/step - loss: 0.3888 - sparse_categorical_accuracy: 0.9250 - val_loss: 0.3516 - val_sparse_categorical_accuracy: 0.8667
    6. Model: "sequential_3"
    7. _________________________________________________________________
    8. Layer (type)                 Output Shape              Param #   
    9. =================================================================
    10. dense_13 (Dense)             (None, 3)                 15        
    11. =================================================================
    12. Total params: 15
    13. Trainable params: 15
    14. Non-trainable params: 0
    15. _________________________________________________________________

    (5)类class搭建神经网络

    用class类封装一个神经网络结构

    class MyModel(Model) model=MyModel

    1. class MyModel(Model):
    2.     def __init__(self):
    3.         super(MyModel,self).__init__()
    4.         # 定义网络结构模块
    5.     def call(self,x):
    6.         # 调用网络结构模块，实现前向传播
    7.         return y
    8.  
    9. model = MyModel()

    __init__() 定义所需网络结构模块call()写出前向传播

    1. from tensorflow.keras import Model
    2. class IrisModel(Model):
    3.     def __init__(self):
    4.         super(IrisModel,self).__init__()
    5.         self.d1 = Dense(3)
    6.     def call(self,x):
    7.         y = self.d1(x)
    8.         return y
    9. model = IrisModel
    1. import tensorflow as tf
    2. from tensorflow.keras.layers import Dense
    3. from tensorflow.keras import Model
    4. from sklearn import datasets
    5. import numpy as np
    6.  
    7. x_train = datasets.load_iris().data
    8. y_train = datasets.load_iris().target
    9.  
    10. np.random.seed(116)
    11. np.random.shuffle(x_train)
    12. np.random.seed(116)
    13. np.random.shuffle(y_train)
    14. tf.random.set_seed(116)
    15.  
    16. class IrisModel(Model):
    17.     def __init__(self):
    18.         super(IrisModel, self).__init__()
    19.         self.d1 = Dense(3, activation='softmax', kernel_regularizer=tf.keras.regularizers.l2())
    20.  
    21.     def call(self, x):
    22.         y = self.d1(x)
    23.         return y
    24.  
    25. model = IrisModel()
    26.  
    27. model.compile(optimizer=tf.keras.optimizers.SGD(lr=0.1),
    28.               loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    29.               metrics=['sparse_categorical_accuracy'])
    30.  
    31. model.fit(x_train, y_train, batch_size=32, epochs=500, validation_split=0.2, validation_freq=20)
    32. model.summary()
    1. Epoch 1/500
    2. 4/4 [==============================] - 0s 748us/step - loss: 2.3173 - sparse_categorical_accuracy: 0.3417
    3. Epoch 2/500
    4. 中间运行结果省略
    5. Epoch 500/500
    6. 4/4 [==============================] - 0s 4ms/step - loss: 0.3888 - sparse_categorical_accuracy: 0.9250 - val_loss: 0.3516 - val_sparse_categorical_accuracy: 0.8667
    7. Model: "iris_model_2"
    8. _________________________________________________________________
    9. Layer (type)                 Output Shape              Param #   
    10. =================================================================
    11. dense_14 (Dense)             multiple                  15        
    12. =================================================================
    13. Total params: 15
    14. Trainable params: 15
    15. Non-trainable params: 0
    16. _________________________________________________________________

    MNIST数据集

    提供六万张20 * 20像素点的0~9手写数字图片和标签，用于训练

    提供一万张28 * 28像素点的0~9手写数字图片和标签，用于测试

    导入MNIST数据集

    1. mnist = tf.keras.datasets.mnist
    2. (x_train,y_train),(x_test,y_test)=mnist.load_data()

    作为输入特征，输入神经网络时，将数据拉伸为一维数组

    tf.keras.layers.Flatten()

    把训练集中的第一个样本x_train[0]可视化出来

    1. plt.imshow(x_train[0],cmap="gray") # 绘制灰度图
    2. plt.show

    代码如下：

    1. import tensorflow as tf
    2. from matplotlib import pyplot as plt
    3.  
    4. mnist = tf.keras.datasets.mnist
    5. (x_train, y_train), (x_test, y_test) = mnist.load_data()
    6.  
    7. # 可视化训练集输入特征的第一个元素
    8. plt.imshow(x_train[0], cmap='gray')  # 绘制灰度图
    9. plt.show()
    10.  
    11. # 打印出训练集输入特征的第一个元素
    12. print("x_train[0]:\n", x_train[0])
    13. # 打印出训练集标签的第一个元素
    14. print("y_train[0]:\n", y_train[0])
    15.  
    16. # 打印出整个训练集输入特征形状
    17. print("x_train.shape:\n", x_train.shape)
    18. # 打印出整个训练集标签的形状
    19. print("y_train.shape:\n", y_train.shape)
    20. # 打印出整个测试集输入特征的形状
    21. print("x_test.shape:\n", x_test.shape)
    22. # 打印出整个测试集标签的形状
    23. print("y_test.shape:\n", y_test.shape)

    ​png
    ​

    1. x_train[0]:
    2.  [[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    3.     0   0   0   0   0   0   0   0   0   0]
    4.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    5.     0   0   0   0   0   0   0   0   0   0]
    6.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    7.     0   0   0   0   0   0   0   0   0   0]
    8.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    9.     0   0   0   0   0   0   0   0   0   0]
    10.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    11.     0   0   0   0   0   0   0   0   0   0]
    12.  [  0   0   0   0   0   0   0   0   0   0   0   0   3  18  18  18 126 136
    13.   175  26 166 255 247 127   0   0   0   0]
    14.  [  0   0   0   0   0   0   0   0  30  36  94 154 170 253 253 253 253 253
    15.   225 172 253 242 195  64   0   0   0   0]
    16.  [  0   0   0   0   0   0   0  49 238 253 253 253 253 253 253 253 253 251
    17.    93  82  82  56  39   0   0   0   0   0]
    18.  [  0   0   0   0   0   0   0  18 219 253 253 253 253 253 198 182 247 241
    19.     0   0   0   0   0   0   0   0   0   0]
    20.  [  0   0   0   0   0   0   0   0  80 156 107 253 253 205  11   0  43 154
    21.     0   0   0   0   0   0   0   0   0   0]
    22.  [  0   0   0   0   0   0   0   0   0  14   1 154 253  90   0   0   0   0
    23.     0   0   0   0   0   0   0   0   0   0]
    24.  [  0   0   0   0   0   0   0   0   0   0   0 139 253 190   2   0   0   0
    25.     0   0   0   0   0   0   0   0   0   0]
    26.  [  0   0   0   0   0   0   0   0   0   0   0  11 190 253  70   0   0   0
    27.     0   0   0   0   0   0   0   0   0   0]
    28.  [  0   0   0   0   0   0   0   0   0   0   0   0  35 241 225 160 108   1
    29.     0   0   0   0   0   0   0   0   0   0]
    30.  [  0   0   0   0   0   0   0   0   0   0   0   0   0  81 240 253 253 119
    31.    25   0   0   0   0   0   0   0   0   0]
    32.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0  45 186 253 253
    33.   150  27   0   0   0   0   0   0   0   0]
    34.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0  16  93 252
    35.   253 187   0   0   0   0   0   0   0   0]
    36.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0 249
    37.   253 249  64   0   0   0   0   0   0   0]
    38.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0  46 130 183 253
    39.   253 207   2   0   0   0   0   0   0   0]
    40.  [  0   0   0   0   0   0   0   0   0   0   0   0  39 148 229 253 253 253
    41.   250 182   0   0   0   0   0   0   0   0]
    42.  [  0   0   0   0   0   0   0   0   0   0  24 114 221 253 253 253 253 201
    43.    78   0   0   0   0   0   0   0   0   0]
    44.  [  0   0   0   0   0   0   0   0  23  66 213 253 253 253 253 198  81   2
    45.     0   0   0   0   0   0   0   0   0   0]
    46.  [  0   0   0   0   0   0  18 171 219 253 253 253 253 195  80   9   0   0
    47.     0   0   0   0   0   0   0   0   0   0]
    48.  [  0   0   0   0  55 172 226 253 253 253 253 244 133  11   0   0   0   0
    49.     0   0   0   0   0   0   0   0   0   0]
    50.  [  0   0   0   0 136 253 253 253 212 135 132  16   0   0   0   0   0   0
    51.     0   0   0   0   0   0   0   0   0   0]
    52.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    53.     0   0   0   0   0   0   0   0   0   0]
    54.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    55.     0   0   0   0   0   0   0   0   0   0]
    56.  [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
    57.     0   0   0   0   0   0   0   0   0   0]]
    58. y_train[0]:
    59.  5
    60. x_train.shape:
    61.  (60000, 28, 28)
    62. y_train.shape:
    63.  (60000,)
    64. x_test.shape:
    65.  (10000, 28, 28)
    66. y_test.shape:
    67.  (10000,)

    Sequential实现数字识别训练
    完整代码如下：

    方法1：

    1. import tensorflow as tf
    2.  
    3. mnist = tf.keras.datasets.mnist
    4. (x_train, y_train), (x_test, y_test) = mnist.load_data()
    5. x_train, x_test = x_train / 255.0, x_test / 255.0
    6.  
    7. model = tf.keras.models.Sequential([
    8.     tf.keras.layers.Flatten(),
    9.     tf.keras.layers.Dense(128, activation='relu'),
    10.     tf.keras.layers.Dense(10, activation='softmax')
    11. ])
    12.  
    13. model.compile(optimizer='adam',
    14.               loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    15.               metrics=['sparse_categorical_accuracy'])
    16.  
    17. model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
    18. model.summary()
    1. Epoch 1/5
    2. 1875/1875 [==============================] - 2s 821us/step - loss: 0.2508 - sparse_categorical_accuracy: 0.9294 - val_loss: 0.1320 - val_sparse_categorical_accuracy: 0.9597
    3. Epoch 2/5
    4. 1875/1875 [==============================] - 1s 766us/step - loss: 0.1117 - sparse_categorical_accuracy: 0.9670 - val_loss: 0.1034 - val_sparse_categorical_accuracy: 0.9703
    5. Epoch 3/5
    6. 1875/1875 [==============================] - 1s 786us/step - loss: 0.0759 - sparse_categorical_accuracy: 0.9771 - val_loss: 0.0853 - val_sparse_categorical_accuracy: 0.9746
    7. Epoch 4/5
    8. 1875/1875 [==============================] - 1s 778us/step - loss: 0.0576 - sparse_categorical_accuracy: 0.9829 - val_loss: 0.0758 - val_sparse_categorical_accuracy: 0.9762
    9. Epoch 5/5
    10. 1875/1875 [==============================] - 1s 757us/step - loss: 0.0441 - sparse_categorical_accuracy: 0.9864 - val_loss: 0.0744 - val_sparse_categorical_accuracy: 0.9761
    11. Model: "sequential_4"
    12. _________________________________________________________________
    13. Layer (type)                 Output Shape              Param #   
    14. =================================================================
    15. flatten_5 (Flatten)          (None, 784)               0         
    16. _________________________________________________________________
    17. dense_15 (Dense)             (None, 128)               100480    
    18. _________________________________________________________________
    19. dense_16 (Dense)             (None, 10)                1290      
    20. =================================================================
    21. Total params: 101,770
    22. Trainable params: 101,770
    23. Non-trainable params: 0
    24. _________________________________________________________________

    上边的是使用Sequential方法

    下边用类实现手写数字识别模型训练

    1. import tensorflow as tf
    2. from tensorflow.keras.layers import Dense, Flatten
    3. from tensorflow.keras import Model
    4.  
    5. mnist = tf.keras.datasets.mnist
    6. (x_train, y_train), (x_test, y_test) = mnist.load_data()
    7. x_train, x_test = x_train / 255.0, x_test / 255.0
    8.  
    9.  
    10. class MnistModel(Model):
    11.     def __init__(self):
    12.         super(MnistModel, self).__init__()
    13.         self.flatten = Flatten()
    14.         self.d1 = Dense(128, activation='relu')
    15.         self.d2 = Dense(10, activation='softmax')
    16.  
    17.     def call(self, x):
    18.         x = self.flatten(x)
    19.         x = self.d1(x)
    20.         y = self.d2(x)
    21.         return y
    22.  
    23.  
    24. model = MnistModel()
    25.  
    26. model.compile(optimizer='adam',
    27.               loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    28.               metrics=['sparse_categorical_accuracy'])
    29.  
    30. model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
    31. model.summary()
    1. Epoch 1/5
    2. 1823/1875 [============================>.] - ETA: 0s - loss: 0.2640 - sparse_categorical_accuracy: 0.9253WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0000s vs `on_test_batch_end` time: 0.0010s). Check your callbacks.
    3. 1875/1875 [==============================] - 2s 894us/step - loss: 0.2601 - sparse_categorical_accuracy: 0.9264 - val_loss: 0.1423 - val_sparse_categorical_accuracy: 0.9550
    4. Epoch 2/5
    5. 1875/1875 [==============================] - 1s 735us/step - loss: 0.1130 - sparse_categorical_accuracy: 0.9665 - val_loss: 0.1014 - val_sparse_categorical_accuracy: 0.9699
    6. Epoch 3/5
    7. 1875/1875 [==============================] - 1s 728us/step - loss: 0.0760 - sparse_categorical_accuracy: 0.9772 - val_loss: 0.0878 - val_sparse_categorical_accuracy: 0.9729
    8. Epoch 4/5
    9. 1875/1875 [==============================] - 1s 787us/step - loss: 0.0575 - sparse_categorical_accuracy: 0.9825 - val_loss: 0.0720 - val_sparse_categorical_accuracy: 0.9768
    10. Epoch 5/5
    11. 1875/1875 [==============================] - 1s 732us/step - loss: 0.0440 - sparse_categorical_accuracy: 0.9865 - val_loss: 0.0786 - val_sparse_categorical_accuracy: 0.9764
    12. Model: "mnist_model_3"
    13. _________________________________________________________________
    14. Layer (type)                 Output Shape              Param #   
    15. =================================================================
    16. flatten_6 (Flatten)          multiple                  0         
    17. _________________________________________________________________
    18. dense_17 (Dense)             multiple                  100480    
    19. _________________________________________________________________
    20. dense_18 (Dense)             multiple                  1290      
    21. =================================================================
    22. Total params: 101,770
    23. Trainable params: 101,770
    24. Non-trainable params: 0
    25. _________________________________________________________________

    FASHION数据集

    提供6万张20 * 28像素点的衣裤等图片和标签，用于训练

    提供一万张28 * 28像素点的衣裤等图片和标签，用于测试

    img

    导入FASHION数据集

    1. fashion = tf.keras.datasets.fashion_mnist
    2.  
    3. (x_train,y_train),(x_test,y_test) = fashion.load_data()

    完整代码如下

    (1)Sequential方法

    1. import tensorflow as tf
    2.  
    3. fashion = tf.keras.datasets.fashion_mnist
    4. (x_train, y_train),(x_test, y_test) = fashion.load_data()
    5. x_train, x_test = x_train / 255.0, x_test / 255.0
    6.  
    7. model = tf.keras.models.Sequential([
    8.     tf.keras.layers.Flatten(),
    9.     tf.keras.layers.Dense(128, activation='relu'),
    10.     tf.keras.layers.Dense(10, activation='softmax')
    11. ])
    12.  
    13. model.compile(optimizer='adam',
    14.               loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    15.               metrics=['sparse_categorical_accuracy'])
    16.  
    17. model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
    18. model.summary()
    1. Epoch 1/5
    2. 1875/1875 [==============================] - 2s 908us/step - loss: 0.5022 - sparse_categorical_accuracy: 0.8241 - val_loss: 0.4137 - val_sparse_categorical_accuracy: 0.8531
    3. Epoch 2/5
    4. 1875/1875 [==============================] - 1s 764us/step - loss: 0.3777 - sparse_categorical_accuracy: 0.8642 - val_loss: 0.4052 - val_sparse_categorical_accuracy: 0.8574
    5. Epoch 3/5
    6. 1875/1875 [==============================] - 1s 733us/step - loss: 0.3383 - sparse_categorical_accuracy: 0.8766 - val_loss: 0.3890 - val_sparse_categorical_accuracy: 0.8609
    7. Epoch 4/5
    8. 1875/1875 [==============================] - 1s 751us/step - loss: 0.3157 - sparse_categorical_accuracy: 0.8838 - val_loss: 0.3711 - val_sparse_categorical_accuracy: 0.8637
    9. Epoch 5/5
    10. 1875/1875 [==============================] - 1s 731us/step - loss: 0.2976 - sparse_categorical_accuracy: 0.8903 - val_loss: 0.3641 - val_sparse_categorical_accuracy: 0.8667
    11. Model: "sequential_5"
    12. _________________________________________________________________
    13. Layer (type)                 Output Shape              Param #   
    14. =================================================================
    15. flatten_7 (Flatten)          (None, 784)               0         
    16. _________________________________________________________________
    17. dense_19 (Dense)             (None, 128)               100480    
    18. _________________________________________________________________
    19. dense_20 (Dense)             (None, 10)                1290      
    20. =================================================================
    21. Total params: 101,770
    22. Trainable params: 101,770
    23. Non-trainable params: 0
    24. _________________________________________________________________

    (2)类方法

    1. import tensorflow as tf
    2. from tensorflow.keras.layers import Dense, Flatten
    3. from tensorflow.keras import Model
    4.  
    5. fashion = tf.keras.datasets.fashion_mnist
    6. (x_train, y_train),(x_test, y_test) = fashion.load_data()
    7. x_train, x_test = x_train / 255.0, x_test / 255.0
    8.  
    9.  
    10. class MnistModel(Model):
    11.     def __init__(self):
    12.         super(MnistModel, self).__init__()
    13.         self.flatten = Flatten()
    14.         self.d1 = Dense(128, activation='relu')
    15.         self.d2 = Dense(10, activation='softmax')
    16.  
    17.     def call(self, x):
    18.         x = self.flatten(x)
    19.         x = self.d1(x)
    20.         y = self.d2(x)
    21.         return y
    22.  
    23.  
    24. model = MnistModel()
    25.  
    26. model.compile(optimizer='adam',
    27.               loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    28.               metrics=['sparse_categorical_accuracy'])
    29.  
    30. model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
    31. model.summary()
    1. Epoch 1/5
    2. 1875/1875 [==============================] - 2s 831us/step - loss: 0.4998 - sparse_categorical_accuracy: 0.8241 - val_loss: 0.4051 - val_sparse_categorical_accuracy: 0.8543
    3. Epoch 2/5
    4. 1875/1875 [==============================] - 1s 738us/step - loss: 0.3737 - sparse_categorical_accuracy: 0.8652 - val_loss: 0.4044 - val_sparse_categorical_accuracy: 0.8572
    5. Epoch 3/5
    6. 1875/1875 [==============================] - 1s 726us/step - loss: 0.3347 - sparse_categorical_accuracy: 0.8785 - val_loss: 0.3784 - val_sparse_categorical_accuracy: 0.8629
    7. Epoch 4/5
    8. 1875/1875 [==============================] - 1s 701us/step - loss: 0.3126 - sparse_categorical_accuracy: 0.8857 - val_loss: 0.3707 - val_sparse_categorical_accuracy: 0.8645
    9. Epoch 5/5
    10. 1875/1875 [==============================] - 1s 780us/step - loss: 0.2967 - sparse_categorical_accuracy: 0.8917 - val_loss: 0.3506 - val_sparse_categorical_accuracy: 0.8698
    11. Model: "mnist_model_4"
    12. _________________________________________________________________
    13. Layer (type)                 Output Shape              Param #   
    14. =================================================================
    15. flatten_8 (Flatten)          multiple                  0         
    16. _________________________________________________________________
    17. dense_21 (Dense)             multiple                  100480    
    18. _________________________________________________________________
    19. dense_22 (Dense)             multiple                  1290      
    20. =================================================================
    21. Total params: 101,770
    22. Trainable params: 101,770
    23. Non-trainable params: 0
    24. _________________________________________________________________
  • 相关阅读:
    【SQL】指定日期的产品价格（IFNULL函数）
    mysql的变量
    Kettle【实践 08】全国地铁线路信息及线路站点座标数据获取脚本及技巧说明（云资源分享:完整ktr脚本及20221008最新数据SQL）
    2023年奔走的总结---吉特日化MES 制药项目 篇二
    【翻译】Efficient Data Loader for Fast Sampling-Based GNN Training on Large Graphs
    stable diffusion实践操作-黑白稿线稿上色
    【微信小程序】6天精准入门（第5天:利用案例与后台的数据交互）附源码
    Lumos学习王佩丰Excel第四讲：排序与选择
    牛客网刷题 | BC117 逆序输出
    三道有趣的算法题
  • 原文地址：https://blog.csdn.net/ks2686/article/details/126594386