How do you verify if a training curve is overfitting?

Question

Eberechi ICHI 2022-11-7

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1845768-how-do-you-verify-if-a-training-curve-is-overfitting

评论： Eberechi ICHI 2022-11-8

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

John D'Errico 2022-11-7

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1845768-how-do-you-verify-if-a-training-curve-is-overfitting#answer_1093673

This is why you need to segregate some of your data for validation purposes.

Split your data into two parts, perhaps selected randomly. For the model to your training data. What was the error in the fit? Note that.

Then test the model on the validation data. If the validation set has an error of roughly the same size as the training set, then you are not overfitting.

However, if the prediction error on the validation set is significantly larger than the training set, then you are overfitting your data.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Eberechi ICHI 2022-11-8

Yes i did, and this is one of the curve for the model. I was wondering if this is overffitting. Although the error curve for the training and validation converges but he training and validation curve above seems to be divergent. Tjhat is actually my concern. Thanks

请先登录，再进行评论。