Keras神经网络全连接层多分类问题的训练损失、验证损失和训练精度、验证精度

灰太狼 2023-06-17 10:53 13阅读 0赞

# AI：Keras神经网络全连接层多分类问题的训练损失、验证损失和训练精度、验证精度，Python #

import keras
    from keras.layers import Dense
    from keras import models
    from keras import optimizers
    
    from sklearn.datasets.samples_generator import make_blobs
    from sklearn.preprocessing import MinMaxScaler
    import matplotlib.pyplot as plt
    
    
    def dummy_data(size, center_size):
        # 制造一批训练数据。
        x, y = make_blobs(n_samples=size, centers=center_size, n_features=2, random_state=1)
    
        scaler = MinMaxScaler()
        # MinMaxScaler：归一到 [ 0，1 ]
        scaler.fit(x)
        x = scaler.transform(x)
    
        # print("原始训练数据")
        # for i in range(len(x)):
        #    print("x=%s 标签:%s" % (x[i], y[i]))
        # print("-----")
    
        plt.scatter(x[:, 0], x[:, 1], c=y)
        plt.grid()
        plt.show()
    
        return x, y
    
    
    def fig_loss(history):
        history_dict = history.history
        loss_values = history_dict['loss']
        val_loss_values = history_dict['val_loss']
        epochs = range(1, len(loss_values) + 1)
        plt.plot(epochs, loss_values, 'b-.', label='Training loss')
        plt.plot(epochs, val_loss_values, 'r--', label='Validation loss')
        plt.title('Training and validation loss')
        plt.xlabel('Epochs')
        plt.ylabel('Loss')
        plt.legend()
        plt.grid()
        plt.show()
    
    
    def fig_acc(history):
        history_dict = history.history
        acc = history_dict['accuracy']
        val_acc = history_dict['val_accuracy']
        epochs = range(1, len(acc) + 1)
        plt.plot(epochs, acc, 'g:', label='Training acc')
        plt.plot(epochs, val_acc, 'r-', label='Validation acc')
        plt.title('Training and validation accuracy')
        plt.xlabel('Epochs')
        plt.ylabel('Accuracy')
        plt.legend()
        plt.grid()
        plt.show()
    
    
    def predict(network):
        print('泛化预测新数据:')
        x_new, y_new = dummy_data(20, 4)
        y_predict = network.predict_classes(x_new)
    
        for i in range(len(x_new)):
            print("数据=%s 正确分类:%s 预测分类:%s" % (x_new[i], y_new[i], y_predict[i]))
    
    
    def dense():
        MY_LAYER_NAME = 'layer_1'
        MY_UNIT = 4
        MY_SHAPE = (2,)
    
        network = models.Sequential()
    
        layer1 = Dense(name=MY_LAYER_NAME, units=MY_UNIT * 2, activation='relu', input_shape=MY_SHAPE)
        network.add(layer1)
    
        layer2 = Dense(units=MY_UNIT, activation="softmax")
        network.add(layer2)
    
        network.compile(loss="categorical_crossentropy",
                        optimizer=optimizers.RMSprop(lr=0.001),
                        metrics=['accuracy'])
    
        # 训练数据
        x_train, y_train = dummy_data(3000, MY_UNIT)
        y_train = keras.utils.to_categorical(y_train)
    
        # 验证数据
        x_val, y_val = dummy_data(800, MY_UNIT)
        y_val = keras.utils.to_categorical(y_val)
    
        my_fit_history = network.fit(x_train, y_train, epochs=60, verbose=1, validation_data=(x_val, y_val))
    
        # 绘制训练损失和验证损失
        fig_loss(my_fit_history)
    
        # 绘制训练精度和验证精度
        fig_acc(my_fit_history)
    
        x_test, y_test = dummy_data(100, 4)
        results = network.evaluate(x_test, keras.utils.to_categorical(y_test))
        print('训练结束,评估模型->')
        print('损失:')
        print(results[0])
    
        print('精度:')
        print(results[1])
    
        # 利用训练好的模型泛化预测前所未知的新数据分类
        predict(network)
    
    
    if __name__ == "__main__":
        dense()

程序实现意图：生成四种聚类数据散点，然后训练这批数据。训练完成后用模型泛化预测新的数据散点。

预训练数据集：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70][]

用来验证的数据集：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 1][]

训练过程中的训练损失和验证损失：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 2][]

训练精度和验证精度：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 3][]

代码运行后的部分输出：

训练结束,评估模型->
    损失:
    0.04747089475393295
    精度:
    0.9900000095367432
    泛化预测新数据:
    数据=[0.5797192  0.50899869] 正确分类:3 预测分类:3
    数据=[0.33139875 0.47067588] 正确分类:3 预测分类:3
    数据=[0.97472153 0.86329803] 正确分类:0 预测分类:0
    数据=[0.92136652 0.84507397] 正确分类:0 预测分类:0
    数据=[1.        0.9613595] 正确分类:0 预测分类:0
    数据=[0.44120969 0.01902758] 正确分类:2 预测分类:2
    数据=[0.36449655 0.58074881] 正确分类:3 预测分类:3
    数据=[0.37128437 0.45202284] 正确分类:3 预测分类:3
    数据=[0.17903116 0.42564499] 正确分类:1 预测分类:1
    数据=[0.81515505 0.98983219] 正确分类:0 预测分类:0
    数据=[0.         0.47412934] 正确分类:1 预测分类:1
    数据=[0.10218844 0.43170872] 正确分类:1 预测分类:1
    数据=[0.29921329 0.00683701] 正确分类:2 预测分类:2
    数据=[0.19979183 0.30467477] 正确分类:1 预测分类:1
    数据=[0.34966961 0.        ] 正确分类:2 预测分类:2
    数据=[0.33670437 0.11068337] 正确分类:2 预测分类:2
    数据=[0.872512 1.      ] 正确分类:0 预测分类:0
    数据=[0.41415172 0.38596844] 正确分类:3 预测分类:3
    数据=[0.08299585 0.32143688] 正确分类:1 预测分类:1
    数据=[0.29881198 0.04069237] 正确分类:2 预测分类:2

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20191128172141198.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/20191128172202966.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 2]: https://img-blog.csdnimg.cn/20191128172237989.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 3]: https://img-blog.csdnimg.cn/20191128172313226.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly96aGFuZ3BoaWwuYmxvZy5jc2RuLm5ldA==,size_16,color_FFFFFF,t_70