Monitoring AI solutions Questions
Practice questions for Monitoring AI solutions topic in Google Professional Machine Learning Engineer. 27 questions covering this domain.
A deployed model has access to its original training data, and the team wants to detect whether production feature distributions are deviating from th...
In Vertex AI Model Monitoring v1, which distance metrics are used by default to compare feature distributions for categorical and numerical features?
Which statement correctly distinguishes Vertex AI Model Monitoring v1 from v2?
A production model is still serving traffic, but the team wants to periodically evaluate it against new labeled data and retrain if quality declines. ...
Which action is explicitly recommended in Vertex AI generative AI responsible AI guidance?
A monitoring team wants to detect when the importance of a key feature changes over time even if the raw input distribution shift is subtle. Which Mod...
A practitioner sees that monitoring detects significant feature skew between training and online serving for a numeric feature. Which root-cause inves...
Which type of model degradation occurs when the relationship between inputs and the target label changes over time?
Which Vertex AI capability captures explanations attributing a prediction to its input features for individual requests?
Which Vertex AI Model Monitoring v2 objective compares production prediction output distributions to a baseline, regardless of input drift?
A team wants alerts in Cloud Monitoring when a Vertex AI endpoint's prediction error rate or latency crosses a threshold. Which approach is recommende...
A monitoring service must capture a sampled subset of online prediction requests and responses for ongoing analysis. Which Vertex AI feature supports ...
A responsible AI review of a generative model requires assessing whether outputs contain harmful content categories during ongoing operations. Which V...
A team wants to configure Vertex AI Model Monitoring v1 to send an email alert when a feature distribution deviation exceeds the threshold. Which conf...
What is the key difference between data drift and concept drift in a deployed model?
A team monitors a deployed classification model and wants to set quality thresholds that trigger retraining when the model's F1 score on periodically ...
A team uses Cloud Logging to capture all errors from a Vertex AI model serving endpoint and wants to be alerted when the error rate exceeds 5% over a ...
In SHAP-based explanations, what does a large positive SHAP value for a feature indicate?
A team wants to configure sampling rate and monitoring frequency for a Vertex AI Model Monitoring v1 job. Which configuration parameter controls how o...
Which Vertex AI explainability method is specifically designed for image models and produces region-level attribution maps by combining Integrated Gra...
Sign in to see all 27 questions
Create a free account to browse all questions — completely free during our launch phase.