Glossary

Unsupervised as monitoring

Use anomaly detection to monitor deployed models. If the distribution of incoming data shifts (more anomalies than usual), it may signal concept drift — the model's predictions may be degrading.

Learn More

Related Terms