Interview: model trained in region A must now serve region B — approach?
hardAnswer
- (1) Diagnose: KS + PSI feature-by-feature; MMD/classifier-drift for multivariate.
- (2) If severe drift, retrain on B data (labeled ideally).
- (3) Domain adaptation: (a) importance weighting via density ratio, (b) adversarial (DANN), (c) CORAL alignment of covariances, (d) fine-tune on small B-labeled set.
- (4) Semi-supervised: use B unlabeled + A labeled.
- (5) Test-time adaptation (BN statistics update on B).
- (6) Monitor performance in B before full rollout.
Check yourself — multiple choice
- Same model
- Diagnose drift (KS/MMD) + retrain if possible + domain adaptation (importance weight / DANN / CORAL / fine-tune small set) + semi-sup + TTA + monitor A→B rollout
- Random
- Not real
Domain adaptation A→B: diagnose + retrain / DANN / CORAL / semi-sup + monitor.
#interview#applications#anomaly-detection
Practise Unsupervised Learning
214 interview questions in this topic.
Related questions
- Interview: 'design an unsupervised fraud detection system'.
- Interview: production fraud detector — daily volume 10M, current FN too high.
- Interview: 'design anomaly detection for a factory sensor with 200 signals.'
- Interview: 'you're asked to segment 10M customers — how?'
- Anomaly detection in time series — what changes?
- How is drift detection an unsupervised problem?