初学keras

在python中安装tensorflow

使用阿里源镜像安装

pip install tensorflow -i https://mirrors.aliyun.com/pypi/simple/

导入 Fashion MNIST 数据集

导入keras数据

import tensorflow as tf
from tensorflow import keras
fashion_mnist = keras.datasets.fashion_mnist
(train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data()

归一化

train_images = train_images / 255.0
test_images = test_images / 255.0

建立模型

输入层，添加层和输出层

model = keras.Sequential([
    keras.layers.Flatten(input_shape=(28, 28)),
    keras.layers.Dense(128, activation='relu'),
    keras.layers.Dense(10)
])

编译模型

model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

训练模型

history =model.fit(train_images, train_labels,epochs=100)

评估模型

test_loss, test_acc = model.evaluate(test_images,  test_labels, verbose=2)

保存模型

mp = "G://机器学习1/iris_model2.h5"
model.save(mp)

由于每次操作都要训练一次模型，我采用保存模型的方式

安装keras

pip install keras -i https://mirrors.aliyun.com/pypi/simple/

加载模型

mp = "G://机器学习1/iris_model1.h5"
model = load_model(mp)

使用模型进行预测

(train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data()
predictions = model.predict(test_images)

建立一个列表保存测试失败的数据

error = []
for i in range(len(predictions)):
    if np.argmax(predictions[i]) != test_labels[i]:
        error.append(i)

对测试集错误的前100个图形查看

for i in range(100):
    plt.subplot(10, 10, i + 1)
    plt.grid(False)
    plt.imshow(test_images[error[i]], cmap=plt.cm.binary)
    plt.xlabel((np.argmax(predictions[error[i]]),test_labels[error[i]]))
plt.show()