A researcher applies SMOTE (Synthetic Minority Over-sampling Technique) to address class imbalance before training. What synthetic minority samples does SMOTE generate?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — SMOTE creates synthetic minority samples by picking a minority observation, finding one of its k nearest minority neighbors, and interpolating a new point along the line segment between them. You're not just copying data or guessing from a distribution; you're building new minority points in the feature space.
Full Explanation
The interpolated-samples approach is correct because SMOTE generates synthetic minority observations by selecting a random minority sample, choosing one of its k nearest minority-class neighbors, and interpolating a new point along the line segment between them: new_sample = minority_obs + rand * (neighbor - minority_obs). The exact-copies-with-jitter approach is not SMOTE; it is more similar to random oversampling with noise. The Gaussian-sampling approach describes a different generative method rather than SMOTE's neighbor-based interpolation. The label-flipping approach would corrupt the data by assigning incorrect labels.