Keras进度栏中的精度是什么意思?

时间:2018-09-28 16:07:46

标签: tensorflow machine-learning keras deep-learning

在Keras中,您会看到类似这样的内容:

{'id': '2'}
{'id': '3'}

但是我很喜欢它:

  

model.fit(X_train,oh_y_train,             batch_size = 512,             纪元= 1,             verbose = 1)

Epoch 1/1 60000/60000 [==============================] - 297s 5ms/step - loss: 0.7048 - acc: 0.7669 60000/60000 [==============================] - 179s 3ms/step Training Set: acc: 94.60% 10000/10000 [==============================] - 30s 3ms/step Test Set: acc: 95.10% 方法中没有验证数据,从第1阶段起测量准确性是什么?

最终的准确性有所不同。

2 个答案:

答案 0 :(得分:3)

这只是训练集的准确性,验证准确性将为val_acc

答案 1 :(得分:1)

此模型适合训练集,然后根据训练集中的观察值验证预测。

所以您在看how good does this regression equation predict the same data it was fit on

当您获得针对保留样本(AKA测试集)的准确性得分时,您会看到how good does this regression equation predict on cases that were not in the training set

过拟合时,在前一种情况下,您的准确性得分会更高。