support@Free-Braindumps.com
Register Login
Cisco
CompTIA
ISC
EC-Council
EXIN
Fortinet
ITIL
Juniper
Microsoft
VMware
Avaya
SAP
Google
NVIDIA
Salesforce
Amazon
Palo Alto Networks
ISC2
CrowdStrike
Anthropic
All Vendors
  1. Home
  2. Exams
  3. Google
  4. PROFESSIONAL MACHINE LEARNING ENGINEER

Free PROFESSIONAL MACHINE LEARNING ENGINEER Exam Braindumps
Professional Machine Learning Engineer Questions 169–176 (page: 23)

Updated On: 7-Jun-2026
Page 23 of 44
PDF with 402 Questions

You are developing a model to detect fraudulent credit card transactions. You need to prioritize detection, because missing even one fraudulent transaction could severely impact the credit card holder. You used AutoML to train a model on users' profile information and credit card transaction data. After training the initial model, you notice that the model is failing to detect many fraudulent transactions. How should you adjust the training parameters in AutoML to improve model performance? (Choose two.)

  1. Increase the score threshold
  2. Decrease the score threshold.
  3. Add more positive examples to the training set
  4. Add more negative examples to the training set
  5. Reduce the maximum number of node hours for training

Answer(s): B,C

Explanation:

Option B is correct because lowering the score threshold makes the model classify more transactions as fraudulent, increasing recall and reducing false negatives (missed frauds). Option C is correct because adding more positive examples (fraud cases) helps the model learn the fraud pattern, improving detection. Incorrect:
A) Increase the score threshold reduces sensitivity, worsening detection of frauds.
D) Adding more negative examples biases the model toward non-fraud, lowering fraud detection.
E) Reducing node hours limits training resources and may underfit; not a targeted improvement for detection.



You need to deploy a scikit-learn classification model to production. The model must be able to serve requests
, and you expect millions of requests per second to the production application from 8 am to 7 pm. You need to minimize the cost of deployment. What should you do?

  1. Deploy an online Vertex AI prediction endpoint. Set the max replica count to 1
  2. Deploy an online Vertex AI prediction endpoint. Set the max replica count to 100
  3. Deploy an online Vertex AI prediction endpoint with one GPU per replica. Set the max replica count to 1
  4. Deploy an online Vertex AI prediction endpoint with one GPU per replica. Set the max replica count to 100

Answer(s): B

Explanation:

Option B is correct because online Vertex AI prediction endpoints scale horizontally to handle high request volume; setting max replicas to 100 provides ample concurrent capacity to meet millions of requests per second from 8 am to 7 pm while minimizing per-request latency and avoiding overprovisioning with GPUs.
A) Incorrect — max replicas=1 would become a bottleneck under high throughput, causing latency and QoS issues.
C) Incorrect — GPUs are unnecessary for scikit-learn inference and raise cost; single-GPU per replica with 1 replica still limits throughput.
D) Incorrect — GPUs together with 100 replicas would overprovision cost for a non-GPU workload, increasing expense without needed acceleration.



You work with a team of researchers to develop state-of-the-art algorithms for financial analysis. Your team develops and debugs complex models in TensorFlow. You want to maintain the ease of debugging while also reducing the model training time. How should you set up your training environment?

  1. Configure a v3-8 TPU VM. SSH into the VM to train and debug the model.
  2. Configure a v3-8 TPU node. Use Cloud Shell to SSH into the Host VM to train and debug the model.
  3. Configure a n1 -standard-4 VM with 4 NVIDIA P100 GPUs. SSH into the VM and use ParameterServerStraregv to train the model.
  4. Configure a n1-standard-4 VM with 4 NVIDIA P100 GPUs. SSH into the VM and use MultiWorkerMirroredStrategy to train the model.

Answer(s): D

Explanation:

Option D is correct because MultiWorkerMirroredStrategy provides synchronous data-parallel training across multiple GPUs, enabling faster training and effective debugging in a unified TF distribution on a single VM. It leverages TensorFlow’s standard distributed APIs without bespoke orchestration, suitable for a team using TensorFlow and needing scalable training with GPU acceleration.
A) Incorrect — TPU v3-8 is not a standard setup for 4 NVIDIA P100 GPUs and SSH-based debugging is not the recommended debugging workflow for distributed TensorFlow; risks mismatched environments.
B) Incorrect — TPU node/Cloud Shell SSH does not align with GPU-based TF distribution on a single VM; TPU setup is different from GPU training.
C) Incorrect — ParameterServerStrategy is older and more complex for synchronous training; MultiWorkerMirroredStrategy is typically preferred for GPU-based multi-worker training.



You created an ML pipeline with multiple input parameters. You want to investigate the tradeoffs between different parameter combinations. The parameter options are:
Input dataset



Max tree depth of the boosted tree regressor



Optimizer learning rate



You need to compare the pipeline performance of the different parameter combinations measured in F1 score, time to train, and model complexity. You want your approach to be reproducible, and track all pipeline runs on the same platform. What should you do?

  1. 1. Use BigQueryML to create a boosted tree regressor, and use the hyperparameter tuning capability.
    2. Configure the hyperparameter syntax to select different input datasets, max tree depths, and optimizer learning rates. Choose the grid search option.
  2. 1. Create a Vertex AI pipeline with a custom model training job as part of the pipeline. Configure the pipeline's parameters to include those you are investigating.
    2. In the custom training step, use the Bayesian optimization method with F1 score as the target to maximize.
  3. 1. Create a Vertex AI Workbench notebook for each of the different input datasets.

    2. In each notebook, run different local training jobs with different combinations of the max tree depth and optimizer learning rate parameters.
    3. After each notebook finishes, append the results to a BigQuery table.
  4. 1. Create an experiment in Vertex AI Experiments.
    2. Create a Vertex AI pipeline with a custom model training job as part of the pipeline. Configure the pipeline's parameters to include those you are investigating.
    3. Submit multiple runs to the same experiment, using different values for the parameters.

Answer(s): D

Explanation:

Option D is correct because Vertex AI Experiments (now part of Vertex AI) provides reproducible tracking of multiple runs with different hyperparameters, metrics (like F1), and artifacts in a centralized way, enabling compare-and-contrast across configurations. It supports logging, lineage, and aggregation across experiments for time, accuracy, and model complexity.
A is incorrect because BigQueryML boosted trees with hyperparameter tuning does not integrate the end-to-end reproducible experiment tracking and multi-metric comparison across pipelines in a single managed experiment workflow.
B is incorrect because while a Vertex AI pipeline with a custom training job enables parameterization, it does not by itself guarantee centralized, reproducible experiment tracking across multiple runs and metrics.
C is incorrect because Workbench notebooks alone do not provide centralized experiment tracking or scalable, repeatable evaluation across multiple parameter configurations.



You received a training-serving skew alert from a Vertex AI Model Monitoring job running in production. You retrained the model with more recent training data, and deployed it back to the Vertex AI endpoint, but you are still receiving the same alert. What should you do?

  1. Update the model monitoring job to use a lower sampling rate.
  2. Update the model monitoring job to use the more recent training data that was used to retrain the model.
  3. Temporarily disable the alert. Enable the alert again after a sufficient amount of new production traffic has passed through the Vertex AI endpoint.
  4. Temporarily disable the alert until the model can be retrained again on newer training data. Retrain the model again after a sufficient amount of new production traffic has passed through the Vertex AI endpoint.

Answer(s): B

Explanation:

Option B is correct because model monitoring alerts tied to a training-serving skew should reflect the data distribution used for training. Updating the monitoring setup to use the most recent training data aligns the baseline with current production inputs, ensuring skew detection thresholds are relevant to the retrained model.
A) Incorrect — sampling rate changes do not address data distribution misalignment between training data and live inputs; they affect detection granularity, not the underlying skew basis.
C) Incorrect — disabling alerts hides issues and does not fix data drift or skew; it postpones necessary validation.
D) Incorrect — delaying retraining or alerting postpones resolution; the right action is aligning monitoring with the latest training data rather than delaying fixes.



You developed a custom model by using Vertex AI to forecast the sales of your company's products based on historical transactional data. You anticipate changes in the feature distributions and the correlations between the features in the near future. You also expect to receive a large volume of prediction requests. You plan to use Vertex AI Model Monitoring for drift detection and you want to minimize the cost. What should you do?

  1. Use the features for monitoring. Set a monitoring-frequency value that is higher than the default.
  2. Use the features for monitoring. Set a prediction-sampling-rate value that is closer to 1 than 0.
  3. Use the features and the feature attributions for monitoring. Set a monitoring-frequency value that is lower than the default.
  4. Use the features and the feature attributions for monitoring. Set a prediction-sampling-rate value that is closer to 0 than 1.

Answer(s): D

Explanation:

Option D is correct because Vertex AI Model Monitoring supports monitoring of features and feature attributions, and lowering the prediction-sampling-rate (closer to 0) reduces monitoring cost while still providing drift signals under high prediction volume. This aligns with anticipating distribution changes and high request throughput.
A is incorrect because monitoring features alone without including feature attributions omits potentially richer drift signals; also increasing monitoring frequency raises cost, opposite of the cost-minimization goal.
B is incorrect because a higher prediction-sampling-rate (closer to 1) increases monitoring cost, not minimizes it, and it does not address feature-attribution monitoring.
C is incorrect because lowering monitoring frequency reduces data collected and may miss drift signals during distribution shifts; including feature attributions is beneficial, but the risk of missed signals remains.



You have recently trained a scikit-learn model that you plan to deploy on Vertex AI. This model will support both online and batch prediction. You need to preprocess input data for model inference. You want to package the model for deployment while minimizing additional code. What should you do?

  1. 1. Upload your model to the Vertex AI Model Registry by using a prebuilt scikit-ieam prediction container.
    2. Deploy your model to Vertex AI Endpoints, and create a Vertex AI batch prediction job that uses the instanceConfig.instanceType setting to transform your input data.
  2. 1. Wrap your model in a custom prediction routine (CPR). and build a container image from the CPR local model.
    2. Upload your scikit learn model container to Vertex AI Model Registry.
    3. Deploy your model to Vertex AI Endpoints, and create a Vertex AI batch prediction job
  3. 1. Create a custom container for your scikit learn model.
    2. Define a custom serving function for your model.
    3. Upload your model and custom container to Vertex AI Model Registry.
    4. Deploy your model to Vertex AI Endpoints, and create a Vertex AI batch prediction job.
  4. 1. Create a custom container for your scikit learn model.
    2. Upload your model and custom container to Vertex AI Model Registry.
    3. Deploy your model to Vertex AI Endpoints, and create a Vertex AI batch prediction job that uses the instanceConfig.instanceType setting to transform your input data.

Answer(s): B

Explanation:

Option B is correct because using a custom prediction routine (CPR) encapsulates preprocessing and inference logic in a single container, minimizing extra code and enabling consistent deployment for both online endpoints and batch predictions in Vertex AI. It allows a unified serving function across offline and real-time use cases without modifying the model artifacts.
A) Incorrect — prebuilt scikit-learn containers do not automatically wrap preprocessing; CPR is required to bundle preprocessing logic.
C) Incorrect — creating a custom container is valid but does not inherently include a CPR pipeline; it omits the standardized inference routine.
D) Incorrect — duplicate option with the same reasoning as C; without CPR, you miss streamlined integration for both prediction modes.



You work for a food product company. Your company's historical sales data is stored in BigQuery.You need to use Vertex AI's custom training service to train multiple TensorFlow models that read the data from BigQuery and predict future sales. You plan to implement a data preprocessing algorithm that performs min-max scaling and bucketing on a large number of features before you start experimenting with the models. You want to minimize preprocessing time, cost, and development effort. How should you configure this workflow?

  1. Write the transformations into Spark that uses the spark-bigquery-connector, and use Dataproc to preprocess the data.
  2. Write SQL queries to transform the data in-place in BigQuery.
  3. Add the transformations as a preprocessing layer in the TensorFlow models.
  4. Create a Dataflow pipeline that uses the BigQueryIO connector to ingest the data, process it, and write it back to BigQuery.

Answer(s): B

Explanation:

Option B is correct because performing min-max scaling and bucketing directly in BigQuery leverages its serverless, scalable SQL engine to preprocess large-scale features with minimal development effort and cost, and the results can be consumed by Vertex AI custom training without extra ETL steps. Incorrect — A: Introducing Spark on Dataproc adds operational overhead and cost for a relatively simple preprocessing task, and may not integrate as cleanly with Vertex AI data ingestion. Incorrect — C: Embedding preprocessing as a TensorFlow layer ties preprocessing to each model, increases duplication, and adds training-time overhead rather than one-off, scalable preprocessing. Incorrect — D: Dataflow offers streaming/ batch processing but adds unnecessary latency and complexity when in-place BigQuery SQL preprocessing suffices.



  PDF
Viewing page 23 of 44
Viewing questions 177 - 184 out of 402 questions


Post your Comments and Discuss Google PROFESSIONAL MACHINE LEARNING ENGINEER exam prep with other Community members:

AI Tutor AI Tutor 👋 I’m here to help!