Drift and model fit · Question 29 of 60
How is overfitting tested for?
1 answer is correct. Medium
Show answer
Answer
- B Performance is evaluated on a separate test dataset not used in training, including rare examples
Explanation
To detect overfitting, you test on separate test data that the model has never seen and that also contains unusual cases. If it performs markedly worse there than on the validation data, it has probably memorized rather than generalized.