A company needs to ensure its machine learning model is not biased against any demographic group. What approach should they take first?
AIncrease dataset size
BBias detection techniques
CModel optimization
DFeature selection
Explanation
Bias detection is crucial for addressing demographic bias before model adjustments.
Q342
You are configuring a machine learning model in Vertex AI. What happens if you set a higher learning rate?
AFaster convergence towards minimum
BIncreased model stability
CRisk of overshooting minima
DReduces accuracy significantly
Explanation
A higher learning rate can cause the model to overshoot minima, leading to instability.
Q343
Which service in Google Cloud is primarily used for deploying machine learning models?
ACloud AI Platform
BBigQuery
CCloud Functions
DCloud Storage
Explanation
Cloud AI Platform is designed for model deployment; BigQuery is for data analysis, Cloud Functions executes code, and Cloud Storage is for file storage.
Q344
A company needs to build a recommendation system. Which Google Cloud service would best provide real-time predictions?
AFirestore
BCloud Run
CBigQuery ML
DAI Platform Predictions
Explanation
AI Platform Predictions allows real-time predictions, while Firestore and Cloud Run are not designed for machine learning predictions, and BigQuery ML is suited for batch predictions.
Q345
What happens when a deployed model on AI Platform is hit with an unexpected spike in requests?
AIt crashes immediately
BIt scales up automatically
CIt produces errors
DIt queues requests indefinitely
Explanation
AI Platform scales horizontally to handle traffic spikes; it does not crash or queue indefinitely and handles errors based on implementation.
Q346
Which service is best for large-scale data preprocessing?
ACloud Dataflow
BCloud Functions
CCloud Storage
DBigQuery
Explanation
Cloud Dataflow efficiently processes and transforms large data sets, while others are less suitable for preprocessing.
Q347
A company needs to implement AutoML but has limited development resources. What is the best approach?
AUse AI Platform's pre-trained models
BHire many data scientists
CBuild everything from scratch
DSpecify a custom ML model
Explanation
Using pre-trained models is effective and saves resources compared to building or customizing models.
Q348
You are configuring an AI model training job in Google Cloud. What happens when using insufficient training data?
AThe model locks up
BThe model overfits
CTraining fails instantly
DThe model has lower cost
Explanation
Insufficient data often leads to overfitting, as the model learns noise instead of patterns.
Q349
Which service is optimal for training machine learning models at scale?
AGoogle Kubernetes Engine
BAI Platform Training
CCloud Functions
DBigQuery ML
Explanation
AI Platform Training is designed specifically for ML model training, while the others serve different purposes.
Q350
A company needs to serve real-time predictions to users. Which GCP service should they use?
ACloud Run
BAI Platform Prediction
CCloud Pub/Sub
DCloud Storage
Explanation
AI Platform Prediction allows real-time predictions, while the other options do not focus on model serving.