记录我学习TensorFlow2时遇到的问题与解决方法。学习可以看coursera上的TensorFlow in Practice 专项课程（学习中）。
0 安装TF2
TensorFlow 2的安装还是比较方便的，直接使用pip安装就可以了。最好使用虚拟python环境，与系统python隔离开来。虚拟环境有pyenv、virtualenv和Docker选择自己喜欢的用就好了。官方的安装教程写的很详细了，跟着操作就不会出问题。官方安装链接
$ pip install --upgrade pip
$ pip install tensorflow
注意：系统需要64位（都是血和泪），python>3.5。
1 初体验
安装好后，可以先跑个MNIST例子试试安装是否正确：
import tensorflow as tf
mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10, activation='softmax')
])
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
model.fit(x_train, y_train, epochs=5)
model.evaluate(x_test, y_test, verbose=2)
运行结果：
Epoch 1/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.3005 - accuracy: 0.9135
Epoch 2/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.1435 - accuracy: 0.9584
Epoch 3/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.1079 - accuracy: 0.9673
Epoch 4/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.0870 - accuracy: 0.9729
Epoch 5/5
1875/1875 [==============================] - 3s 2ms/step - loss: 0.0747 - accuracy: 0.9770
313/313 - 0s - loss: 0.0746 - accuracy: 0.9763
如果顺利就会输出上面结果，虽然现在还不知道在干什么。
我遇到的问题
由于网络问题mnist.load_data()无法下载MNIST 数据集。解决方法，将数据集下载下来链接，从本地加载数据集。
import numpy as np
def load_data(path):
with np.load(path, allow_pickle=True) as f:
x_train, y_train = f['x_train'], f['y_train']
x_test, y_test = f['x_test'], f['y_test']
return (x_train, y_train), (x_test, y_test)
(x_train, y_train), (x_test, y_test) = load_data("./datasets/mnist.npz")
2 深度学习基础
什么是机器学习？
我的理解是计算机从提供的数据中总结经验（模型），并利用经验对未知的数据进行预测。
Tom Mitchell provides a more modern definition: “A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.”