Overfitting & Underfitting
A model can learn its training data too well and then fail on new data. This is called overfitting. In this lesson you see it happen, learn how to notice it, and learn the fixes.
New words
- Training error
- How wrong the model is on the data it learned from.
- Test error
- How wrong the model is on new data it has never seen.
- Underfitting
- The model is too simple, so it is wrong even on training data.
- Overfitting
- The model memorises the training data, noise included, and fails on new data.
- Bias and variance
- Bias is a too-simple model. Variance is a model that changes too much with the data.
- Regularisation
- A penalty that stops the model from becoming too complicated.
- Validation set
- A part of the data used to choose settings, before the final test.
What you will learn
- Why can a model be too good?
- What are training error and test error?
- What is underfitting?
- What is a good fit, and what is overfitting?
- What is the bias and variance idea?
- How do you notice overfitting?
- How do you fix overfitting?
- Try it: slide the flexibility
- Common mistakes
- Explain it to a friend
- Key takeaways
10 quiz questions at the end.
The full lesson, with interactive demos and the quiz, opens after you sign in with Google. It is free.