DNN模型训练
代码：

使用 fit 方法使模型对训练数据拟合

model.fit(x_train, y_train,
batch_size=128,
epochs=10,
verbose=1)
输出：
Epoch 1/10
60000/60000 [==============================] - 7s 114us/sample - loss: 0.2281 - acc: 0.9327s - loss: 0.2594 - acc: 0. - ETA: 1s - loss: 0.2535 - acc: 0.9 - ETA: 1s - loss:
Epoch 2/10
60000/60000 [==============================] - 8s 129us/sample - loss: 0.0830 - acc: 0.9745s - loss: 0.0814 - ac
Epoch 3/10
60000/60000 [==============================] - 8s 127us/sample - loss: 0.0553 - acc: 0.9822
Epoch 4/10
60000/60000 [==============================] - 7s 117us/sample - loss: 0.0397 - acc: 0.9874s - los
Epoch 5/10
60000/60000 [==============================] - 8s 129us/sample - loss: 0.0286 - acc: 0.9914
Epoch 6/10
60000/60000 [==============================] - 8s 136us/sample - loss: 0.0252 - acc: 0.9919
Epoch 7/10
60000/60000 [==============================] - 8s 129us/sample - loss: 0.0204 - acc: 0.9931s - lo
Epoch 8/10
60000/60000 [==============================] - 8s 135us/sample - loss: 0.0194 - acc: 0.9938
Epoch 9/10
60000/60000 [==============================] - 7s 109us/sample - loss: 0.0162 - acc: 0.9948
Epoch 10/10
60000/60000 [==============================] - ETA: 0s - loss: 0.0149 - acc: 0.994 - 7s 117us/sample - loss: 0.0148 - acc: 0.9948
其中epoch表示批次，表示将全量的数据迭代10次。