Course covers what is overfitting and how to address it. But does not talk about how to identify it. Is it done through plotting? Are there other techniques?
The fitting occurs in the training phase of the development. After training, you need to test your model. If your test score is much worse than your training score (e.g., during training, your model gets 99% correct, but during testing, it only scores 90%), then it is quite likely that during training your model overfitted the data.
Hi @mohammedsuhail.k good question. The denition itself contains the answer. Overfit means that you fit your training data too well that is not able to generalize. By saving valudation and testing data you will see if you overfit or not
Hello, @mohammedsuhail.k,
The lesson in C2 W3 “Bias and variance” has a series of lecture videos on overfitting. Particularly, you will see how to identify it in the lecture “learning curves” when it talks about “High variance”.
Cheers,
Raymond
1 Like