这是过度拟合吗? [英] Is this overfitting?
问题描述
我的CNN在训练数据上的表现很好(准确度为96%,丢失1.%),但在测试数据上的表现却很差(准确度为50%,丢失3.5%).
I have a CNN that is performing very well (96% accuracy, 1.~ loss) on training data but poorly (50% accuracy, 3.5 loss) on testing data.
推荐答案
过拟合的特征签名是指验证损失开始增加,而训练损失则继续减少,即:
The telltale signature of overfitting is when your validation loss starts increasing, while your training loss continues decreasing, i.e.:
(图片摘自Wikipedia上的过度拟合)
(Image adapted from Wikipedia entry on overfitting)
还有其他一些图表明过度拟合(源):
Here are some other plots indicating overfitting (source):
另请参见SO线程如何知道是否拟合不足还是过度拟合?.
很明显,您的损失图确实表现出了这种行为,所以是的,您确实过拟合了.
Clearly, your loss plot does exhibit such behavior, so yes, you are indeed overfitting.
相反,您在评论中链接到的图:
On the contrary, the plot you have linked to in a comment:
不会表现出这种行为,因此在这里并不是实际过拟合(您刚刚达到饱和点,超过此点验证错误就不会进一步改善).
does not exhibit such behavior, hence here you are not actually overfitting (you just have reached a saturation point, beyond which your validation error is not further improving).
这篇关于这是过度拟合吗?的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!