求翻译：关于训练次数，很多人都会有认识上的误区，认为训练次数是越多越好。对于训确实练样本来说，训练次数越多，拟合程度就越高，但是验证样本的误差并不一定会越来越小。因为当网络训练到一定程度之后会出现过度拟合的现象，即网络对训练样本的拟合度越来越高，但是验证样本的误差曲线保持不变或反而上升。这是因为网络训练首先学习的是训练集中的“共性”，而后才是每个训练样本各自的“特性”，训练过度很容易导致对训练样本的过度匹配现象，因此验证样本的误差会随着训练次数的增加而变大。是什么意思？

正在翻译，请等待...

正在翻译，请等待...

About the training number of times, very many people all can have in the understanding erroneous zone, thought the training number of times is more the better.Regarding teaches practices the sample truly, the training number of times are more, the fitting degree is higher, but the confirmation sampl

On training times, many people will understand the misunderstanding, that training is the number, the more the better. For training practice samples, train more, the higher degree of fitting, but the validation sample error does not necessarily become smaller and smaller. Because when a network trai