You built a deep learning-based image classification model by using on-premises data. You want to use Vertex AI to deploy the model to production. Due to security concerns, you cannot move your data t

Vertex Explainable AI provides insights into why a model made a specific prediction, but it doesn't automatically monitor for data skew or drift that leads to performance changes over time.

Similar to A, example-based explanations in Vertex Explainable AI help understand individual predictions but do not provide continuous monitoring for shifts in overall data distribution or model performance degradation.

While Vertex AI Model Monitoring can detect training-serving skew, the specific requirement for detecting 'input data distribution might change over time' is more directly addressed by feature drift detection, alongside skew, especially with feature attribution, for a more comprehensive understanding of performance shifts.

Vertex AI Model Monitoring allows you to detect data skew (differences between training and serving data) and data drift (changes in serving data over time), which are critical for identifying model performance degradation due to changing input data distribution. Enabling feature attribution skew and drift detection specifically helps understand if the importance of features is changing, providing deeper insights into model behavior shifts.