Google Professional Machine Learning Engineer

• #21

  You're developing a custom TensorFlow classification model based on tabular data stored in BigQuery. The dataset comprises hundreds of millions of rows with both categorical and numerical features. Your goal is to use a MaxMin scaler on some numerical features and apply one-hot encoding to categorical features like SKU names. The model will be trained over multiple epochs, and you aim to minimize both effort and cost. What approach should you take?

  • AA. 1. Compose a SQL query to generate a separate lookup table for scaling the numerical features. 2. Utilize a TensorFlow-based model from Hugging Face deployed to BigQuery to encode the text features. Direct the resulting BigQuery view into Vertex AI Training.
  • BB. 1. Utilize BigQuery to scale the numerical features. 2. Directly input the features into Vertex AI Training. Allow TensorFlow to handle the one-hot text encoding.
  • CC. 1. Employ TFX components with Dataflow to encode the text features and scale the numerical features. 2. Export the outcomes to Cloud Storage as TFRecords. Input the data into Vertex AI Training.
  • DD. 1. Draft a SQL query to create a separate lookup table for scaling the numerical features. 2. Perform the one-hot text encoding in BigQuery. Direct the resulting BigQuery view into Vertex AI Training.
  Show answer
• #22

  You need to train an object detection model to identify bounding boxes around Post-it Notes® in an image. Post-it Notes can have a variety of background colors and shapes. You have a dataset with 1000 images with a maximum size of 1.4MB and a CSV file containing annotations stored in Cloud Storage. You want to select a training method that reliably detects Post-it Notes of any relative size in the image and that minimizes the time to train a model. What should you do?

  • AA. Use the Cloud Vision API in Vertex AI with OBJECT_LOCALIZATION type, and filter the detected objects that match the Post-it Note category only.
  • BB. Upload your dataset into Vertex AI. Use Vertex AI AutoML Vision Object Detection with accuracy as the optimization metric, early stopping enabled, and no training budget specified.
  • CC. Write a Python training application that trains a custom vision model on the training set. Autopackage the application, and configure a custom training job in Vertex AI.
  • DD. Write a Python training application that performs transfer learning on a pre-trained neural network. Autopackage the application, and configure a custom training job in Vertex AI.
  Show answer
• #23

  You are employed at a prominent healthcare company, tasked with creating advanced algorithms for a range of applications. Your dataset consists of unstructured text data with specialized annotations. Your objective is to extract and categorize different medical expressions with these annotations. What course of action should you take?

  • AA. Utilize the Healthcare Natural Language API for medical entity extraction.
  • BB. Employ a BERT-based model to fine-tune a medical entity extraction model.
  • CC. Utilize Vertex AI AutoML Entity Extraction to train a medical entity extraction model.
  • DD. Develop a customized medical entity extraction model using TensorFlow.
  Show answer
• #24

  You are employed by a major retailer and have received a request to categorize your customers based on their buying patterns. The buying records of all customers have been uploaded to BigQuery. You have a hunch that there might be multiple distinct customer segments, but you're uncertain about the exact number and the shared characteristics among them. Your goal is to discover the most efficient approach. What steps should you take?

  • AA. Create a k-means clustering model using BigQuery ML. Allow BigQuery to automatically optimize the number of clusters.
  • BB. Create a new dataset in Dataprep that references your BigQuery table. Use Dataprep to identify similarities within each column.
  • CC. Use the Data Labeling Service to label each customer record in BigQuery. Train a model on your labeled data using AutoML Tables. Review the evaluation metrics to understand whether there is an underlying pattern in the data.
  • DD. Get a list of the customer segments from your company’s Marketing team. Use the Data Labeling Service to label each customer record in BigQuery according to the list. Analyze the distribution of labels in your dataset using Data Studio.
  Show answer
• #25

  As an ML engineer at a manufacturing company, you're currently working on a predictive maintenance project. The goal is to create a classification model that predicts whether a critical machine will experience a failure within the next three days. This predictive capability allows the repair team to address potential issues before they lead to a breakdown. While routine maintenance for the machine is cost-effective, a failure can result in significant expenses. You've trained multiple binary classifiers to make predictions about the machine's failure, where a prediction of 1 indicates the model foresees a failure. Now, during the evaluation phase on a separate dataset, you face the decision of selecting a model that emphasizes detection. However, you also need to ensure that over 50% of the maintenance tasks initiated by your model are genuinely related to impending machine failures. Which model should you opt for to achieve this balance?

  • AA. The model with the highest area under the receiver operating characteristic curve (AUC ROC) and precision greater than 0.5
  • BB. The model with the lowest root mean squared error (RMSE) and recall greater than 0.5.
  • CC. The model with the highest recall where precision is greater than 0.5.
  • DD. The model with the highest precision where recall is greater than 0.5.
  Show answer
• #26

  You are a member of the data science team at a multinational beverage company. Your task is to create an ML model for predicting the profitability of a new line of naturally flavored bottled waters in various locations. You have access to historical data containing information such as product types, product sales volumes, expenses, and profits for all regions. What should you select as the input and output variables for your model?

  • AA. Use latitude, longitude, and product type as features. Use profit as model output.
  • BB. Use latitude, longitude, and product type as features. Use revenue and expenses as model outputs.
  • CC. Use product type and the feature cross of latitude with longitude, followed by binning, as features. Use profit as model output.
  • DD. Use product type and the feature cross of latitude with longitude, followed by binning, as features. Use revenue and expenses as model outputs.
  Show answer
• #27

  You have developed a Python module using Keras to train a regression model with two architectures: linear regression and deep neural network (DNN). The module utilizes the training_method argument to select the architecture, and for the DNN, it includes learning_rate and num_hidden_layers as hyperparameters. You plan to employ Vertex AI's hyperparameter tuning service with a budget of 100 trials to determine the optimal model architecture and hyperparameter values that minimize training loss and enhance performance. How should you proceed?

  • AA. Run a single hyperparameter tuning job for 100 trials. Set num_hidden_layers as a conditional hyperparameter based on its parent hyperparameter training_method, and set learning_rate as a non-conditional hyperparameter.
  • BB. Conduct two separate hyperparameter tuning jobs: one for linear regression with 50 trials and another for DNN with 50 trials. Compare their performance on a common validation set and select the hyperparameters yielding the lowest training loss.
  • CC. Execute one hyperparameter tuning job with training_method as the hyperparameter for 50 trials. Choose the architecture with the lowest training loss, then further tune it and its corresponding hyperparameters for an additional 50 trials.
  • DD. Run a single hyperparameter tuning job for 100 trials. Set both num_hidden_layers and learning_rate as conditional hyperparameters based on their parent hyperparameter training_method.
  Show answer
• #28

  You are part of a data center team responsible for server maintenance. Your management has tasked you with developing a predictive maintenance solution using monitoring data to detect potential server failures. However, the incident data has not been labeled yet. What should be your initial step in this process?

  • AA. Train a time-series model to predict the machines’ performance values. Configure an alert if a machine’s actual performance values significantly differ from the predicted performance values.
  • BB. Develop a simple heuristic (e.g., based on z-score) to label the machines’ historical performance data. Use this heuristic to monitor server performance in real time.
  • CC. Hire a team of qualified analysts to review and label the machines’ historical performance data. Train a model based on this manually labeled dataset.
  • DD. Develop a simple heuristic (e.g., based on z-score) to label the machines’ historical performance data. Train a model to predict anomalies based on this labeled dataset.
  Show answer
• #29

  You're employed at a credit card company and have been assigned the task of developing a custom fraud detection model using Vertex AI AutoML Tables, leveraging historical data. Your primary goal is to enhance the detection of fraudulent transactions while keeping false positives to a minimum. What optimization objective should you select when training the model?

  • AA. An optimization objective that minimizes Log loss
  • BB. An optimization objective that maximizes the Precision at a Recall value of 0.50
  • CC. An optimization objective that maximizes the area under the precision-recall curve (AUC PR) value
  • DD. An optimization objective that maximizes the area under the receiver operating characteristic curve (AUC ROC) value
  Show answer
• #30

  You're involved in a project that utilizes Neural Networks. The dataset you've been given contains columns with varying ranges. During the data preparation process for model training, you observe that gradient optimization is struggling to converge to a favorable solution. What is the recommended action to take?

  • AA. Use feature construction to combine the strongest features.
  • BB. Use the representation transformation (normalization) technique.
  • CC. Improve the data cleaning step by removing features with missing values.
  • DD. Change the partitioning step to reduce the dimension of the test set and have a larger training set.
  Show answer
• #31

  You work for a large retailer, and you need to build a model to predict customer churn. The company has a dataset of historical customer data, including customer demographics purchase history, and website activity. You need to create the model in BigQuery ML and thoroughly evaluate its performance. What should you do?

  • AA. Create a linear regression model in BigQuery ML, and register the model in Vertex AI Model Registry. Evaluate the model performance in Vertex AI.
  • BB. Create a logistic regression model in BigQuery ML and register the model in Vertex AI Model Registry. Evaluate the model performance in Vertex AI.
  • CC. Create a linear regression model in BigQuery ML. Use the ML.EVALUATE function to evaluate the model performance.
  • DD. Create a logistic regression model in BigQuery ML. Use the ML.CONFUSION_MATRIX function to evaluate the model performance.
  Show answer
• #32

  You are training models in Vertex AI using data that spans across multiple Google Cloud projects. You need to find, track, and compare the performance of the different versions of your models. Which Google Cloud services should you include in your ML workflow?

  • AA. Dataplex, Vertex AI Feature Store, and Vertex AI TensorBoard
  • BB. Vertex AI Pipelines, Vertex AI Feature Store, and Vertex AI Experiments
  • CC. Dataplex, Vertex AI Experiments, and Vertex AI ML Metadata
  • DD. Vertex AI Pipelines, Vertex AI Experiments, and Vertex AI Metadata
  Show answer
• #33

  You are a member of the operations team at a global company that oversees a substantial fleet of on-premises servers situated in a handful of data centers worldwide. Your team is tasked with gathering monitoring data from these servers, which includes details on CPU and memory usage. In the event of a server incident, your team is responsible for resolving the issue. However, incident data has not yet been appropriately labeled. Your management has requested the development of a predictive maintenance solution that utilizes VM monitoring data to identify potential failures and subsequently notifies the service desk team. What should be your initial step in this process?

  • AA. Train a time-series model to predict the machines’ performance values. Configure an alert if a machine’s actual performance values significantly differ from the predicted performance values.
  • BB. Implement a simple heuristic (e.g., based on z-score) to label the machines’ historical performance data. Train a model to predict anomalies based on this labeled dataset.
  • CC. Develop a simple heuristic (e.g., based on z-score) to label the machines’ historical performance data. Test this heuristic in a production environment.
  • DD. Hire a team of qualified analysts to review and label the machines’ historical performance data. Train a model based on this manually labeled dataset.
  Show answer
• #34

  You trained a model for sentiment analysis in TensorFlow Keras, saved it in SavedModel format, and deployed it with Vertex AI Predictions as a custom container. You selected a random sentence from the test set, and used a REST API call to send a prediction request. The service returned the error: “Could not find matching concrete function to call loaded from the SavedModel. Got: Tensor("inputs:0", shape=(None,), dtype=string). Expected: TensorSpec(shape=(None, None), dtype=tf.int64, name='inputs')”. You want to update the model’s code and fix the error while following Google-recommended best practices. What should you do?

  • AA. Combine all preprocessing steps in a function, and call the function on the string input before requesting the model’s prediction on the processed input.
  • BB. Combine all preprocessing steps in a function, and update the default serving signature to accept a string input wrapped into the preprocessing function call.
  • CC. Create a custom layer that performs all preprocessing steps, and update the Keras model to accept a string input followed by the custom preprocessing layer.
  • DD. Combine all preprocessing steps in a function, and update the Keras model to accept a string input followed by a Lambda layer wrapping the preprocessing function.
  Show answer
• #35

  You've been tasked with operationalizing a proof-of-concept ML model developed with Keras. This model was trained in a Jupyter notebook on a data scientist's local machine, which includes data validation and model analysis cells. Your goal is to automate and orchestrate these notebook steps for weekly retraining, considering an anticipated increase in training data volume. To optimize cost-efficiency and leverage managed services, what steps should you take?

  • AA. Move the Jupyter notebook to a Notebooks instance on the largest N2 machine type, and schedule the execution of the steps in the Notebooks instance using Cloud Scheduler.
  • BB. Write the code as a TensorFlow Extended (TFX) pipeline orchestrated with Vertex AI Pipelines. Use standard TFX components for data validation and model analysis, and use Vertex AI Pipelines for model retraining.
  • CC. Rewrite the steps in the Jupyter notebook as an Apache Spark job, and schedule the execution of the job on ephemeral Dataproc clusters using Cloud Scheduler.
  • DD. Extract the steps contained in the Jupyter notebook as Python scripts, wrap each script in an Apache Airflow BashOperator, and run the resulting directed acyclic graph (DAG) in Cloud Composer.
  Show answer
• #36

  You work as an ML engineer at a travel company, and you've been studying customers' travel behavior for an extended period. During this time, you've deployed models to forecast customers' vacation patterns. You've noticed that customers' vacation choices are influenced by seasonality and holidays, and these seasonal patterns remain consistent across different years. Your goal is to efficiently store and compare model versions and performance metrics across multiple years. How should you approach this task?

  • AA. Store the performance statistics in Cloud SQL. Query that database to compare the performance statistics across the model versions.
  • BB. Create versions of your models for each season per year in Vertex AI. Compare the performance statistics across the models in the Evaluate tab of the Vertex AI UI.
  • CC. Store the performance statistics of each pipeline run in Kubeflow under an experiment for each season per year. Compare the results across the experiments in the Kubeflow UI.
  • DD. Store the performance statistics of each version of your models using seasons and years as events in Vertex ML Metadata. Compare the results across the slices.
  Show answer
• #37

  You are developing a classification model to support predictions for your company’s various products. The dataset you were given for model development has class imbalance You need to minimize false positives and false negatives What evaluation metric should you use to properly train the model?

  • AA. F1 score
  • BB. Recall
  • CC. Accuracy
  • DD. Precision
  Show answer
• #38

  You recently developed a regression model based on a training dataset that does not contain personally identifiable information (PII) data in compliance with regulatory requirements. Before deploying the model, you perform post-training analysis on multiple data slices and discover that the model is under-predicting for users who are more than 60 years old. You want to remove age bias while maintaining similar training offline performance. What should you do?

  • AA. Perform correlation analysis on the training feature set against the age column, and remove features that are highly correlated with age from the training and evaluation sets.
  • BB. Review the data distribution for each feature against the bucketized age column for the training and evaluation sets, and introduce preprocessing to even irregular feature distributions.
  • CC. Split the training and evaluation sets for users below and above 60 years old, and train one specialized model for each user group.
  • DD. Apply a calibration layer at post-processing that matches the prediction distributions of users below and above 60 years old.
  Show answer
• #39

  You are employed at a biotech startup focused on experimenting with deep learning ML models inspired by biological organisms. Your team frequently engages in early-stage experiments involving novel ML model architectures and develops custom TensorFlow operations in C++. Training your models involves large datasets and substantial batch sizes, with a typical batch comprising 1024 examples, each approximately 1 MB in size. Furthermore, the average size of a network, including all weights and embeddings, is 20 GB. In light of these requirements, which hardware should you select for your models?

  • AA. A cluster with 2 n1-highcpu-64 machines, each with 8 NVIDIA Tesla V100 GPUs (128 GB GPU memory in total), and a n1-highcpu-64 machine with 64 vCPUs and 58 GB RAM
  • BB. A cluster with 2 a2-megagpu-16g machines, each with 16 NVIDIA Tesla A100 GPUs (640 GB GPU memory in total), 96 vCPUs, and 1.4 TB RAM
  • CC. A cluster with an n1-highcpu-64 machine with a v2-8 TPU and 64 GB RAM
  • DD. A cluster with 4 n1-highcpu-96 machines, each with 96 vCPUs and 86 GB RAM
  Show answer
• #40

  You are employed by a prominent social network service provider where users publish articles and engage in news discussions. With millions of comments posted daily and over 200 human moderators screening comments for appropriateness, your team is developing an ML model to assist these human moderators in content review. The model assigns scores to each comment and identifies suspicious ones for human review. Which metric(s) should be employed to monitor the model's performance?

  • AA. Number of messages flagged by the model per minute.
  • BB. Number of messages flagged by the model per minute confirmed as being inappropriate by humans.
  • CC. Precision and recall estimates based on a random sample of 0.1% of raw messages each minute sent to a human for review.
  • DD. Precision and recall estimates based on a sample of messages flagged by the model as potentially inappropriate each minute.
  Show answer

Want to track your score, take timed mock exams and get AI explanations? Create a free account