SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-early-stopping

With early stopping, when do you end training?

When loss on the validation set starts to rise.

Start playing free

  1. When training loss reaches exactly zero
  2. When validation loss starts to increase ✔
  3. After exactly 100 epochs

Why: Training loss may keep falling, but once validation loss turns upward, the model is starting to overfit. Stopping there is a simple form of regularization.

Huh, didn't know that: Early stopping ends training before the training loss has finished decreasing, on purpose.

The dad joke
I use early stopping at dinner. When dessert loss starts rising, I put the fork down.
Like this card?
Play the game

References

Answer: When validation loss starts to increase. Training loss may keep falling, but once validation loss turns upward, the model is starting to overfit. Stopping there is a simple form of regularization.

  1. Google Machine Learning Glossary: early stopping developers.google.com/machine-learning/glossary
    you intentionally stop training the model when the loss on a validation dataset starts to increase
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceWhich algorithm makes gradient descent practical for multilayer neural networks?Artificial IntelligenceWhat happens to layers whose training gradients vanish?Artificial IntelligenceWhat is the main idea of mixed-precision neural network training?

That's stuff everyone should know.