0
$\begingroup$

On the y-axis you've got RMSE and on the x-axis you've got the number of epochs. Then in blue, the validation error, in red the training error.

What do you think is the optimal number of epochs before the model starts overfitting? I would go with 74 epochs.

enter image description here

Improve this question
$\endgroup$
2
  • 2
    $\begingroup$ If a model performs better on the data used to train it than the data used to validate it, it is overfitting. As to what number of epochs you should use for your data and model, we cannot answer this. Perhaps you should look into regularization first, but it is hard to say without any information on your data or model. $\endgroup$
    – Frans Rodenburg
    Commented Jun 14, 2019 at 5:26
  • $\begingroup$ Somebody answered me this, regarding the same post: "Best checkpoint would be the one where validation loss stops decreasing further and remains constant. In your case as you said it would be around epoch 70. Or you can add early stopping in your algorithm so that model itself would take care of the point where it needs to stop training further." $\endgroup$
    – Stephen
    Commented Jun 14, 2019 at 5:42

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.