Google Professional Machine Learning Engineer

• #101

  You are a member of the AI team at an automotive company, and your current project involves building a visual defect detection model using TensorFlow and Keras. To enhance the performance of your model, you intend to integrate various image augmentation techniques, including translation, cropping, and contrast adjustments. These augmentation methods will be applied randomly to each training batch. Your objective is to optimize the data processing pipeline for both runtime efficiency and efficient utilization of computational resources. What steps should you take to achieve this goal?

  • AA. Embed the augmentation functions dynamically in the tf.Data pipeline.
  • BB. Embed the augmentation functions dynamically as part of Keras generators.
  • CC. Use Dataflow to create all possible augmentations, and store them as TFRecords.
  • DD. Use Dataflow to create the augmentations dynamically per training run, and stage them as TFRecords.
  Show answer
• #102

  You have created an extensive neural network using TensorFlow Keras, and it's anticipated to require several days for training. The model exclusively relies on TensorFlow's native operations and conducts training with high-precision arithmetic. Your objective is to enhance the code to enable distributed training through tf.distribute.Strategy. Additionally, you aim to configure an appropriate virtual machine instance within Compute Engine to reduce the overall training duration. What steps should you take to achieve this?

  • AA. Select an instance with an attached GPU, and gradually scale up the machine type until the optimal execution time is reached. Add MirroredStrategy to the code, and create the model in the strategy’s scope with batch size dependent on the number of replicas.
  • BB. Create an instance group with one instance with attached GPU, and gradually scale up the machine type until the optimal execution time is reached. Add TF_CONFIG and MultiWorkerMirroredStrategy to the code, create the model in the strategy’s scope, and set up data autosharing.
  • CC. Create a TPU virtual machine, and gradually scale up the machine type until the optimal execution time is reached. Add TPU initialization at the start of the program, define a distributed TPUStrategy, and create the model in the strategy’s scope with batch size and training steps dependent on the number of TPUs.
  • DD. Create a TPU node, and gradually scale up the machine type until the optimal execution time is reached. Add TPU initialization at the start of the program, define a distributed TPUStrategy, and create the model in the strategy’s scope with batch size and training steps dependent on the number of TPUs.
  Show answer
• #103

  You are logged into the Vertex AI Pipeline UI and noticed that an automated production TensorFlow training pipeline finished three hours earlier than a typical run. You do not have access to production data for security reasons, but you have verified that no alert was logged in any of the ML system’s monitoring systems and that the pipeline code has not been updated recently. You want to debug the pipeline as quickly as possible so you can determine whether to deploy the trained model. What should you do?

  • AA. Navigate to Vertex AI Pipelines, and open Vertex AI TensorBoard. Check whether the training regime and metrics converge.
  • BB. Access the Pipeline run analysis pane from Vertex AI Pipelines, and check whether the input configuration and pipeline steps have the expected values.
  • CC. Determine the trained model’s location from the pipeline’s metadata in Vertex ML Metadata, and compare the trained model’s size to the previous model.
  • DD. Request access to production systems. Get the training data’s location from the pipeline’s metadata in Vertex ML Metadata, and compare data volumes of the current run to the previous run.
  Show answer
• #104

  You are tasked with training a natural language model for text classification, specifically on product descriptions. This dataset comprises millions of examples and contains a vocabulary of 100,000 unique words. Your goal is to preprocess the words individually so that they can be effectively input into a recurrent neural network. What steps should you take to achieve this?

  • AA. Create a hot-encoding of words, and feed the encodings into your model.
  • BB. Identify word embeddings from a pre-trained model, and use the embeddings in your model.
  • CC. Sort the words by frequency of occurrence, and use the frequencies as the encodings in your model.
  • DD. Assign a numerical value to each word from 1 to 100,000 and feed the values as inputs in your model.
  Show answer
• #105

  You are developing a real-time prediction engine that streams files, potentially containing Personally Identifiable Information (PII), to Google Cloud. To scan these files, you plan to use the Cloud Data Loss Prevention (DLP) API. What measures should you implement to guarantee that the PII remains inaccessible to unauthorized persons?

  • AA. Stream all files to Google Cloud, and then write the data to BigQuery. Periodically conduct a bulk scan of the table using the DLP API.
  • BB. Stream all files to Google Cloud, and write batches of the data to BigQuery. While the data is being written to BigQuery, conduct a bulk scan of the data using the DLP API.
  • CC. Create two buckets of data: Sensitive and Non-sensitive. Write all data to the Non-sensitive bucket. Periodically conduct a bulk scan of that bucket using the DLP API, and move the sensitive data to the Sensitive bucket.
  • DD. Create three buckets of data: Quarantine, Sensitive, and Non-sensitive. Write all data to the Quarantine bucket. Periodically conduct a bulk scan of that bucket using the DLP API, and move the data to either the Sensitive or Non-Sensitive bucket.
  Show answer
• #106

  You work for an online retailer. Your company has a few thousand short lifecycle products. Your company has five years of sales data stored in BigQuery. You have been asked to build a model that will make monthly sales predictions for each product. You want to use a solution that can be implemented quickly with minimal effort. What should you do?

  • AA. Use Prophet on Vertex AI Training to build a custom model.
  • BB. Use BigQuery ML to build a statistical ARIMA_PLUS model.
  • CC. Use Vertex AI Forecast to build a NN-based model.
  • DD. Use TensorFlow on Vertex AI Training to build a custom model.
  Show answer
• #107

  You are employed by a magazine distribution company, and your task is to develop a predictive model for identifying customers who will renew their subscriptions for the upcoming year. You have utilized your company's historical data as the training dataset and have built a TensorFlow model, deploying it on Vertex AI. Now, your objective is to identify the most influential customer attribute for each prediction generated by the model. How should you proceed?

  • AA. Stream prediction results to BigQuery. Use BigQuery’s CORR(X1, X2) function to calculate the Pearson correlation coefficient between each feature and the target variable.
  • BB. Use Vertex Explainable AI. Submit each prediction request with the explain keyword to retrieve feature attributions using the sampled Shapley method.
  • CC. Use Vertex AI Workbench user-managed notebooks to perform a Lasso regression analysis on your model, which will eliminate features that do not provide a strong signal.
  • DD. Use the What-If tool in Google Cloud to determine how your model will perform when individual features are excluded. Rank the feature importance in order of those that caused the most significant performance drop when removed from the model.
  Show answer
• #108

  You are employed at a bank, and you've developed a customized model to determine whether a loan application should be flagged for human review. The input features required for this model are stored within a BigQuery table. The model has exhibited strong performance, and you are in the process of preparing it for deployment in a production setting. However, due to compliance requirements, it is now imperative that the model provides explanations for each prediction it makes. Your objective is to incorporate this explanatory capability into your model's code with minimal effort while ensuring that the explanations offered are as accurate as possible. How should you proceed to accomplish this?

  • AA. Create an Vertex AI AutoML tabular model by using the BigQuery data with integrated Vertex Explainable AI.
  • BB. Create a BigQuery ML deep neural network model and use the ML.EXPLAIN_PREDICT method with the num_integral_steps parameter.
  • CC. Upload the custom model to Vertex AI Model Registry and configure feature-based attribution by using sampled Shapley with input baselines.
  • DD. Update the custom serving container to include sampled Shapley-based explanations in the prediction outputs.
  Show answer
• #109

  You are using Keras and TensorFlow to develop a fraud detection model. Records of customer transactions are stored in a large table in BigQuery. You need to preprocess these records in a cost-effective and efficient way before you use them to train the model. The trained model will be used to perform batch inference in BigQuery. How should you implement the preprocessing workflow?

  • AA. Implement a preprocessing pipeline by using Apache Spark, and run the pipeline on Dataproc. Save the preprocessed data as CSV files in a Cloud Storage bucket.
  • BB. Load the data into a pandas DataFrame. Implement the preprocessing steps using pandas transformations, and train the model directly on the DataFrame.
  • CC. Perform preprocessing in BigQuery by using SQL. Use the BigQueryClient in TensorFlow to read the data directly from BigQuery.
  • DD. Implement a preprocessing pipeline by using Apache Beam, and run the pipeline on Dataflow. Save the preprocessed data as CSV files in a Cloud Storage bucket.
  Show answer
• #110

  You are an ML engineer at a retail company. You have built a model that predicts which coupon to offer an ecommerce customer at checkout based on the items in their cart. When a customer goes to checkout, your serving pipeline, which is hosted on Google Cloud, joins the customer's existing cart with a row in a BigQuery table that contains the customers' historic purchase behavior and uses that as the model's input. The web team is reporting that your model is returning predictions too slowly to load the coupon offer with the rest of the web page. How should you speed up your model's predictions?

  • AA. Create a materialized view in BigQuery with the necessary data for predictions.
  • BB. Use a low latency database for the customers’ historic purchase behavior.
  • CC. Deploy your model to more instances behind a load balancer to distribute traffic.
  • DD. Attach an NVIDIA P100 GPU to your deployed model’s instance.
  Show answer
• #111

  You recently trained an XGBoost model using tabular data and plan to make it available as an HTTP microservice for internal use. Anticipating a low volume of incoming requests, you seek the most efficient method to deploy the model with minimal effort and latency. What is your best option?

  • AA. Deploy the model to BigQuery ML using the CREATE MODEL statement with the BOOSTED_TREE_REGRESSOR option, and call the BigQuery API from the microservice.
  • BB. Develop a Flask-based application, package it in a custom container on Vertex AI, and deploy it to Vertex AI Endpoints.
  • CC. Create a Flask-based application, package it in a Docker image, and deploy it to Google Kubernetes Engine using Autopilot mode.
  • DD. Utilize a prebuilt XGBoost Vertex container to create a model, and deploy it to Vertex AI Endpoints.
  Show answer
• #112

  You're an ML engineer in an agricultural research team, focusing on a crop disease detection tool for identifying leaf rust spots in crop images as an indicator of disease presence and severity. These spots exhibit variability in shape and size and are indicative of disease severity levels. Your objective is to create a highly accurate solution for predicting disease presence and severity. What steps should you take?

  • AA. Create an object detection model that can localize the rust spots.
  • BB. Develop an image segmentation ML model to locate the boundaries of the rust spots.
  • CC. Develop a template matching algorithm using traditional computer vision libraries.
  • DD. Develop an image classification ML model to predict the presence of the disease.
  Show answer
• #113

  You work for an online travel agency that also sells advertising placements on its website to other companies. You have been asked to predict the most relevant web banner that a user should see next. Security is important to your company. The model latency requirements are 300ms@p99, the inventory is thousands of web banners, and your exploratory analysis has shown that navigation context is a good predictor. You want to implement the simplest solution. How should you configure the prediction pipeline?

  • AA. Embed the client on the website, and then deploy the model on AI Platform Prediction.
  • BB. Embed the client on the website, deploy the gateway on App Engine, and then deploy the model on AI Platform Prediction.
  • CC. Embed the client on the website, deploy the gateway on App Engine, deploy the database on Cloud Bigtable for writing and for reading the user's navigation context, and then deploy the model on AI Platform Prediction.
  • DD. Embed the client on the website, deploy the gateway on App Engine, deploy the database on Memorystore for writing and for reading the user's navigation context, and then deploy the model on Google Kubernetes Engine.
  Show answer
• #114

  You are employed by an online publisher that distributes news articles to a vast audience of over 50 million readers. As part of your responsibilities, you have developed an AI model designed to make content recommendations for the company's weekly newsletter. A recommendation is deemed successful if the recipient opens the recommended article within two days of the newsletter's publication date and spends at least one minute on the page. To calculate the success metric, you have access to all the necessary data in BigQuery, which is updated on an hourly basis. Your model has been trained using data spanning eight weeks, with the observation that its performance tends to decline below an acceptable baseline after five weeks. Additionally, the model's training process requires 12 hours to complete. Your primary objective is to ensure that the model consistently performs above the acceptable baseline while optimizing operational costs. Given this scenario, what approach should you adopt to monitor the model effectively and determine when it is necessary to initiate retraining?

  • AA. Use Vertex AI Model Monitoring to detect skew of the input features with a sample rate of 100% and a monitoring frequency of two days.
  • BB. Schedule a cron job in Cloud Tasks to retrain the model every week before the newsletter is created.
  • CC. Schedule a weekly query in BigQuery to compute the success metric.
  • DD. Schedule a daily Dataflow job in Cloud Composer to compute the success metric.
  Show answer
• #115

  You've developed an ML model using AI Platform and are now moving it into production. The model currently serves a few thousand queries per second but is facing latency issues. Requests are handled by a load balancer, which distributes them across multiple CPU-only Kubeflow pods on Google Kubernetes Engine (GKE). To enhance serving latency without altering the underlying infrastructure, what steps should you take?

  • AA. Significantly increase the max_batch_size TensorFlow Serving parameter.
  • BB. Switch to the tensorflow-model-server-universal version of TensorFlow Serving.
  • CC. Significantly increase the max_enqueued_batches TensorFlow Serving parameter.
  • DD. Recompile TensorFlow Serving using the source to support CPU-specific optimizations. Instruct GKE to choose an appropriate baseline minimum CPU platform for serving nodes.
  Show answer
• #116

  Your team is tasked with developing a model to predict if images contain a driver’s license, passport, or credit card. The data engineering team has already constructed the pipeline and created a dataset, comprising 10,000 images with driver’s licenses, 1,000 images with passports, and 1,000 images with credit cards. Your objective now is to train a model using the following label map: ['drivers_license', 'passport', 'credit_card']. Which loss function is most suitable for this task?

  • AA. Categorical hinge
  • BB. Binary cross-entropy
  • CC. Categorical cross-entropy
  • DD. Sparse categorical cross-entropy
  Show answer
• #117

  You intend to migrate a scikit-learn classifier model to TensorFlow, planning to train the TensorFlow model using the same training set as the scikit-learn model. Subsequently, you aim to compare their performances using a common test set, logging the evaluation metrics of each model manually using the Vertex AI Python SDK, and comparing them based on their F1 scores and confusion matrices. How should you log the metrics?

  • AA. Utilize the aiplatform.log_classification_metrics function to log the F1 score, and employ the aiplatform.log_metrics function to log the confusion matrix.
  • BB. Utilize the aiplatform.log_classification_metrics function to log both the F1 score and the confusion matrix.
  • CC. Utilize the aiplatform.log_metrics function to log both the F1 score and the confusion matrix.
  • DD. Utilize the aiplatform.log_metrics function to log the F1 score, and employ the aiplatform.log_classification_metrics function to log the confusion matrix.
  Show answer
• #118

  You are analyzing customer data for a healthcare organization that is stored in Cloud Storage. The data contains personally identifiable information (PII). You need to perform data exploration and preprocessing while ensuring the security and privacy of sensitive fields. What should you do?

  • AA. Use the Cloud Data Loss Prevention (DLP) API to de-identify the PII before performing data exploration and preprocessing.
  • BB. Use customer-managed encryption keys (CMEK) to encrypt the PII data at rest, and decrypt the PII data during data exploration and preprocessing.
  • CC. Use a VM inside a VPC Service Controls security perimeter to perform data exploration and preprocessing.
  • DD. Use Google-managed encryption keys to encrypt the PII data at rest, and decrypt the PII data during data exploration and preprocessing.
  Show answer
• #119

  You are employed at a retail company and have access to a managed tabular dataset within Vertex AI, which encompasses sales data from three distinct stores. This dataset incorporates various features, including store names and sale timestamps. Your objective is to leverage this data to train a model capable of making sales predictions for an upcoming new store. To accomplish this, you must divide the data into training, validation, and test sets. What approach should you employ for this data split?

  • AA. Use Vertex AI manual split, using the store name feature to assign one store for each set
  • BB. Use Vertex AI default data split
  • CC. Use Vertex AI chronological split, and specify the sales timestamp feature as the time variable
  • DD. Use Vertex AI random split, assigning 70% of the rows to the training set, 10% to the validation set, and 20% to the test set
  Show answer
• #120

  You are employed by a company that offers an anti-spam service for detecting and concealing spam content on social media platforms. Currently, your company relies on a list of 200,000 keywords to identify potential spam posts. If a post contains a significant number of these keywords, it's marked as spam. You are considering incorporating machine learning to assist in identifying spam posts for human review. What is the primary benefit of introducing machine learning in this business scenario?

  • AA. Posts can be compared to the keyword list much more quickly.
  • BB. New problematic phrases can be identified in spam posts.
  • CC. A much longer keyword list can be used to flag spam posts.
  • DD. Spam posts can be flagged using far fewer keywords.
  Show answer

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