Concept guide · Foundations
What is generalisation in machine learning?
A slower explanation for when the short interview answer is not enough. Start with the intuition, then connect it to the technical idea and the way an interviewer expects you to reason about it.
ML area
Foundations → Generalisation
Interview format
Core
Question
ML-011
Start with the idea
In simple terms
Generalisation means the model keeps working on new examples that resemble the situation where it will actually be used.
Make it concrete
A model that memorises the training customers may score perfectly in development but fail when it sees customers from next month.
Build the answer
How to reason about it
- 1Separate data used to fit the model from data used to judge it.
- 2Make the split reflect time, groups, or other deployment boundaries.
- 3Inspect failures and important slices instead of trusting one average score.
Go one level deeper
The technical explanation
Training performance is not enough because the model has already had the opportunity to adapt to those examples. Held-out validation and test procedures estimate how the model behaves on future data, but the estimate is only meaningful when the split matches deployment. Leakage, distribution shift, repeated tuning on the test set, and a non-representative evaluation population can all make generalisation look better than it is.
Say this in the interview
Generalisation is the ability to make useful predictions on new examples drawn from the deployment situation. A model can fit its training data extremely well and still generalise poorly if it memorises noise, exploits leakage, or faces a different population in production. Generalisation is estimated with data and procedures that were not used to make modelling choices.
Keep preparing