Google Professional Machine Learning Engineer

• #221

  You are currently in the process of training a machine learning model for object detection. Your dataset comprises approximately three million X-ray images, each with an approximate size of 2 GB. You have set up the training process using Vertex AI Training, utilizing a Compute Engine instance equipped with 32 cores, 128 GB of RAM, and an NVIDIA P100 GPU. However, you've observed that the model training process is taking an extended period. Your objective is to reduce the training time without compromising the model's performance. What steps should you take to achieve this?

  • AA. Increase the instance memory to 512 GB and increase the batch size.
  • BB. Replace the NVIDIA P100 GPU with a v3-32 TPU in the training job.
  • CC. Enable early stopping in your Vertex AI Training job.
  • DD. Use the tf.distribute.Strategy API and run a distributed training job.
  Tampilkan jawaban
• #222

  You are employed by a gaming company with millions of customers worldwide. Your games offer a real-time chat feature that enables players to communicate with each other in over 20 languages. These messages are translated in real time using the Cloud Translation API. Your task is to create an ML system that moderates the chat in real time while ensuring consistent performance across various languages, all without altering the serving infrastructure. You initially trained a model using an in-house word2vec model to embed the chat messages translated by the Cloud Translation API. However, this model exhibits notable variations in performance among different languages. How can you enhance the model's performance in this scenario?

  • AA. Add a regularization term such as the Min-Diff algorithm to the loss function.
  • BB. Train a classifier using the chat messages in their original language.
  • CC. Replace the in-house word2vec with GPT-3 or T5.
  • DD. Remove moderation for languages for which the false positive rate is too high.
  Tampilkan jawaban
• #223

  You have developed an application that uses a chain of multiple scikit-learn models to predict the optimal price for your company’s products. The workflow logic is shown in the diagram. Members of your team use the individual models in other solution workflows. You want to deploy this workflow while ensuring version control for each individual model and the overall workflow. Your application needs to be able to scale down to zero. You want to minimize the compute resource utilization and the manual effort required to manage this solution. What should you do?

  • AA. Expose each individual model as an endpoint in Vertex AI Endpoints. Create a custom container endpoint to orchestrate the workflow.
  • BB. Create a custom container endpoint for the workflow that loads each model’s individual files Track the versions of each individual model in BigQuery.
  • CC. Expose each individual model as an endpoint in Vertex AI Endpoints. Use Cloud Run to orchestrate the workflow.
  • DD. Load each model’s individual files into Cloud Run. Use Cloud Run to orchestrate the workflow. Track the versions of each individual model in BigQuery.
  Tampilkan jawaban
• #224

  As an ML engineer at a major grocery retail chain with stores across various regions, you have been tasked with developing an inventory prediction model. The model will incorporate features such as region, store location, historical demand, and seasonal popularity. You intend for the algorithm to update its learning daily based on new inventory data. Which algorithms would be most suitable for constructing this model?

  • AA. Classification
  • BB. Reinforcement Learning
  • CC. Recurrent Neural Networks (RNN)
  • DD. Convolutional Neural Networks (CNN)
  Tampilkan jawaban
• #225

  Your production demand forecasting pipeline preprocesses raw data using Dataflow before model training and prediction. This involves applying Z-score normalization to data in BigQuery and then writing it back. With new training data added weekly, your goal is to enhance efficiency by reducing both computation time and manual effort. What steps should you take to achieve this?

  • AA. Normalize the data using Google Kubernetes Engine.
  • BB. Translate the normalization algorithm into SQL for use with BigQuery.
  • CC. Use the normalizer_fn argument in TensorFlow's Feature Column API.
  • DD. Normalize the data with Apache Spark using the Dataproc connector for BigQuery.
  Tampilkan jawaban
• #226

  You work for a manufacturing company and your task is to train a custom image classification model to detect product defects at the end of an assembly line. Although your model is performing well, some images in your holdout set are consistently mislabeled with high confidence. You want to use Vertex AI to gain insights into your model’s results. What should you do?

  • AA. Configure feature-based explanations by using Integrated Gradients. Set the visualization type to PIXELS, and set the clip_percent_upperbound to 95.
  • BB. Create an index by using Vertex AI Matching Engine. Query the index with your mislabeled images.
  • CC. Configure feature-based explanations by using XRAI. Set the visualization type to OUTLINES, and set the polarity to positive.
  • DD. Configure example-based explanations. Specify the embedding output layer to be used for the latent space representation.
  Tampilkan jawaban
• #227

  You are in the process of implementing a batch inference ML pipeline within Google Cloud. The model, developed using TensorFlow, is stored in SavedModel format within Cloud Storage. Your task involves applying this model to a historical dataset, which comprises a substantial 10 TB of data stored within a BigQuery table. How should you proceed to perform the inference effectively?

  • AA. Export the historical data to Cloud Storage in Avro format. Configure a Vertex AI batch prediction job to generate predictions for the exported data
  • BB. Import the TensorFlow model by using the CREATE MODEL statement in BigQuery ML. Apply the historical data to the TensorFlow model
  • CC. Export the historical data to Cloud Storage in CSV format. Configure a Vertex AI batch prediction job to generate predictions for the exported data
  • DD. Configure a Vertex AI batch prediction job to apply the model to the historical data in BigQuery
  Tampilkan jawaban
• #228

  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 to the cloud. You are aware that the input data distribution might change over time. You need to detect model performance changes in production. What should you do?

  • AA. Use Vertex Explainable AI for model explainability. Configure feature-based explanations.
  • BB. Use Vertex Explainable AI for model explainability. Configure example-based explanations.
  • CC. Create a Vertex AI Model Monitoring job. Enable training-serving skew detection for your model.
  • DD. Create a Vertex AI Model Monitoring job. Enable feature attribution skew and drift detection for your model.
  Tampilkan jawaban
• #229

  You are building a custom image classification model and plan to use Vertex AI Pipelines to implement the end-to-end training. Your dataset consists of images that need to be preprocessed before they can be used to train the model. The preprocessing steps include resizing the images, converting them to grayscale, and extracting features. You have already implemented some Python functions for the preprocessing tasks. Which components should you use in your pipeline?

  • AA. DataprocSparkBatchOp and CustomTrainingJobOp
  • BB. DataflowPythonJobOp, WaitGcpResourcesOp, and CustomTrainingJobOp
  • CC. dsl.ParallelFor, dsl.component, and CustomTrainingJobOp
  • DD. ImageDatasetImportDataOp, dsl.component, and Vertex AI AutoMLImageTrainingJobRunOp
  Tampilkan jawaban
• #230

  You developed a tree model based on an extensive feature set of user behavioral data. The model has been in production for 6 months. New regulations were just introduced that require anonymizing personally identifiable information (PII), which you have identified in your feature set using the Cloud Data Loss Prevention API. You want to update your model pipeline to adhere to the new regulations while minimizing a reduction in model performance. What should you do?

  • AA. Redact the features containing PII data, and train the model from scratch.
  • BB. Mask the features containing PII data, and tune the model from the last checkpoint.
  • CC. Use key-based hashes to tokenize the features containing PII data, and train the model from scratch.
  • DD. Use deterministic encryption to tokenize the features containing PII data, and tune the model from the last checkpoint.
  Tampilkan jawaban
• #231

  You work for a telecommunications company, and your task is to build a model for predicting which customers may fail to pay their next phone bill. The goal is to offer assistance to at-risk customers, such as service discounts and bill deadline extensions. Your dataset in BigQuery includes various features like Customer_id, Age, Salary, Sex, Average bill value, Number of phone calls in the last month, and Average duration of phone calls. Your objective is to address potential bias issues while maintaining model accuracy. What should you do?

  • AA. First, check for meaningful correlations between the sensitive features and other features. Then, train a BigQuery ML boosted trees classification model while excluding both the sensitive features and any significantly correlated features.
  • BB. Train a BigQuery ML boosted trees classification model using all available features. Afterward, use the ML.GLOBAL_EXPLAIN method to compute global attribution values for each feature. If any of the sensitive features have importance values exceeding a specified threshold, exclude those features and retrain the model.
  • CC. Train a BigQuery ML boosted trees classification model with all the features. Next, employ the ML.EXPLAIN_PREDICT method to determine attribution values for each feature per customer in a test set. If, for any individual customer, the importance value of any feature surpasses a predefined threshold, rebuild the model without that feature.
  • DD. Establish a fairness metric based on accuracy across the sensitive features. Train a BigQuery ML boosted trees classification model with all features. Then, utilize the trained model to make predictions on a test set. Join this data back with the sensitive features and compute a fairness metric to assess whether it meets your specified requirements.
  Tampilkan jawaban
• #232

  You trained a model in a Vertex AI Workbench notebook that has good validation RMSE. You defined 20 parameters with the associated search spaces that you plan to use for model tuning. You want to use a tuning approach that maximizes tuning job speed. You also want to optimize cost, reproducibility, model performance, and scalability where possible if they do not affect speed. What should you do?

  • AA. Set up a cell to run a hyperparameter tuning job using Vertex AI Vizier with val_rmse specified as the metric in the study configuration.
  • BB. Using a dedicated Python library such as Hyperopt or Optuna, configure a cell to run a local hyperparameter tuning job with Bayesian optimization.
  • CC. Refactor the notebook into a parametrized and dockerized Python script, and push it to Container Registry. Use the UI to set up a hyperparameter tuning job in Vertex AI. Use the created image and include Grid Search as an algorithm.
  • DD. Refactor the notebook into a parametrized and dockerized Python script, and push it to Container Registry. Use the command line to set up a hyperparameter tuning job in Vertex AI. Use the created image and include Random Search as an algorithm where maximum trial count is equal to parallel trial count.
  Tampilkan jawaban
• #233

  You're employed by an organization running a streaming music service with a custom production model providing "next song" recommendations based on user listening history. The model is deployed on a Vertex AI endpoint and recently retrained with fresh data, showing positive offline test results. Now, you aim to test the new model in production with minimal complexity. What approach should you take?

  • AA. Establish a new Vertex AI endpoint for the updated model and deploy it. Develop a service to randomly direct 5% of production traffic to the new endpoint. Monitor end-user metrics like listening time. Gradually increase traffic to the new endpoint if end-user metrics improve over time compared to the old model.
  • BB. Capture incoming prediction requests in BigQuery and set up an experiment in Vertex AI Experiments. Conduct batch predictions for both models using the captured data. Compare model performance side by side using the user's selected song. Deploy the new model to production if its metrics outperform the previous model.
  • CC. Deploy the new model to the existing Vertex AI endpoint and utilize traffic splitting to direct 5% of production traffic to it. Monitor end-user metrics such as listening time. If the new model outperforms the old one over time, gradually increase the traffic directed to the new model.
  • DD. Configure model monitoring for the existing Vertex AI endpoint to detect prediction drift, setting a threshold for alerts. Update the model on the endpoint from the previous one to the new model. Revert to the previous model if alerted to prediction drift.
  Tampilkan jawaban
• #234

  After deploying several versions of an image classification model on AI Platform, you aim to track and compare their performance over time. What approach should you take to effectively perform this comparison?

  • AA. Compare the loss performance for each model on a held-out dataset.
  • BB. Compare the loss performance for each model on the validation data.
  • CC. Compare the receiver operating characteristic (ROC) curve for each model using the What-If Tool.
  • DD. Compare the mean average precision across the models using the Continuous Evaluation feature.
  Tampilkan jawaban
• #235

  You downloaded a TensorFlow language model pre-trained on a proprietary dataset by another company, and you tuned the model with Vertex AI Training by replacing the last layer with a custom dense layer. The model achieves the expected offline accuracy; however, it exceeds the required online prediction latency by 20ms. You want to optimize the model to reduce latency while minimizing the offline performance drop before deploying the model to production. What should you do?

  • AA. Apply post-training quantization on the tuned model, and serve the quantized model.
  • BB. Use quantization-aware training to tune the pre-trained model on your dataset, and serve the quantized model.
  • CC. Use pruning to tune the pre-trained model on your dataset, and serve the pruned model after stripping it of training variables.
  • DD. Use clustering to tune the pre-trained model on your dataset, and serve the clustered model after stripping it of training variables.
  Tampilkan jawaban
• #236

  You've created a custom model using Vertex AI to predict your company's product sales, relying on historical transactional data. You foresee potential shifts in feature distributions and correlations between these features in the near future. Additionally, you anticipate a significant influx of prediction requests. In light of this, you intend to employ Vertex AI Model Monitoring for drift detection while keeping costs to a minimum. What step should you take to achieve this?

  • AA. Use the features for monitoring. Set a monitoring-frequency value that is higher than the default.
  • BB. Use the features for monitoring. Set a prediction-sampling-rate value that is closer to 1 than 0.
  • CC. Use the features and the feature attributions for monitoring. Set a monitoring-frequency value that is lower than the default.
  • DD. Use the features and the feature attributions for monitoring. Set a prediction-sampling-rate value that is closer to 0 than 1.
  Tampilkan jawaban
• #237

  You work for a hospital aiming to optimize its operation scheduling process. To predict the number of beds needed for patients based on scheduled surgeries, you have one year of data organized in 365 rows, including variables like the number of scheduled surgeries, the number of beds occupied, and the date for each day. Your goal is to maximize the speed of model development and testing. What should you do?

  • AA. Create a BigQuery table. Use BigQuery ML to build a regression model with the number of beds as the target variable and features like the number of scheduled surgeries and date-related features (e.g., day of the week) as predictors.
  • BB. Create a BigQuery table. Use BigQuery ML to build an ARIMA model with the number of beds as the target variable and date as the time variable.
  • CC. Create a Vertex AI tabular dataset. Train an AutoML regression model with the number of beds as the target variable and features like the number of scheduled minor surgeries and date-related features (e.g., day of the week) as predictors.
  • DD. Create a Vertex AI tabular dataset. Train a Vertex AI AutoML Forecasting model with the number of beds as the target variable, the number of scheduled surgeries as a covariate, and date as the time variable.
  Tampilkan jawaban
• #238

  Your company, which specializes in selling corporate electronic products globally, has accumulated substantial historical customer data stored in BigQuery. You need to create a model to predict customer lifetime value (CLTV) over the next three years. The approach should be straightforward and efficient. What should you do?

  • AA. Create a Vertex AI Workbench notebook. Use IPython magic to run the CREATE MODEL statement to create an ARIMA model.
  • BB. Access BigQuery Studio in the Google Cloud console. Run the CREATE MODEL statement in the SQL editor to create an AutoML regression model.
  • CC. Create a Vertex AI Workbench notebook. Use IPython magic to run the CREATE MODEL statement to create an AutoML regression model.
  • DD. Access BigQuery Studio in the Google Cloud console. Run the CREATE MODEL statement in the SQL editor to create an ARIMA model.
  Tampilkan jawaban
• #239

  While developing an image recognition model using PyTorch with the ResNet50 architecture, your code has been successfully tested on a small subsample using your local laptop. However, your full dataset consists of 200,000 labeled images, and you aim to efficiently scale your training workload while keeping costs low. You have access to 4 V100 GPUs. What steps should you take to achieve this?

  • AA. Create a Google Kubernetes Engine cluster with a node pool that has 4 V100 GPUs. Prepare and submit a TFJob operator to this node pool.
  • BB. Create a Vertex AI Workbench user-managed notebooks instance with 4 V100 GPUs, and use it to train your model.
  • CC. Package your code with Setuptools, and use a pre-built container. Train your model with Vertex AI using a custom tier that contains the required GPUs.
  • DD. Configure a Compute Engine VM with all the dependencies that launches the training. Train your model with Vertex AI using a custom tier that contains the required GPUs.
  Tampilkan jawaban
• #240

  You've been assigned the task of deploying prototype code into a production environment. The feature engineering component is written in PySpark and operates on Dataproc Serverless, while model training is conducted using a Vertex AI custom training job. These two steps are currently disjointed, requiring manual execution of model training after the feature engineering phase concludes. Your objective is to establish a scalable and maintainable production workflow that seamlessly connects and tracks these steps. What should you do?

  • AA. Create a Vertex AI Workbench notebook, utilize it to submit the Dataproc Serverless feature engineering job, and then submit the custom model training job within the same notebook. Execute the notebook cells sequentially to link the steps end-to-end.
  • BB. Create a Vertex AI Workbench notebook, initiate an Apache Spark context within the notebook, and execute the PySpark feature engineering code. Additionally, utilize the same notebook to execute the custom model training job in TensorFlow. Run the notebook cells sequentially to interconnect the steps from start to finish.
  • CC. Utilize the Kubeflow pipelines SDK to compose code specifying two components: the first being a Dataproc Serverless component responsible for initiating the feature engineering job, and the second being a custom component wrapped using the create_custom_training_job_from_component utility to launch the custom model training job. Create a Vertex AI Pipelines job to link and execute both of these components.
  • DD. Employ the Kubeflow pipelines SDK to draft code outlining two components: the first component initiates an Apache Spark context to execute the PySpark feature engineering code, and the second component runs the TensorFlow custom model training code. Establish a Vertex AI Pipelines job to interconnect and execute both of these components.
  Tampilkan jawaban

Ingin melacak skor, mengikuti ujian simulasi berwaktu, dan mendapat penjelasan AI? Buat akun gratis