Which pair of techniques is used to guard against overfitting?
Show answer and explanation
Answer: B
Cross-validation rotates the validation fold across the data, so performance is estimated on several unseen partitions rather than one lucky split and the gap to training performance becomes visible early. Regularisation penalises model complexity through L1 or L2 terms, dropout or early stopping, which constrains the model from chasing noise. Together they give both a reliable signal and a mechanism to act on it.
Why the other options are wrong
- A. More layers and more epochs increase capacity and training time, which typically worsens overfitting rather than preventing it.
- C. Fitting one sample repeatedly teaches the model that sample and destroys generalisation.
- D. Removing the test set eliminates the only independent evidence of how the model behaves on unseen data.