You work for a magazine publisher and have been tasked with predicting whether customers will cancel their annual subscription. In your exploratory data analysis, you find that 90% of individuals renew their subscription every year, and only 10% of individuals cancel their subscription. After training a NN Classifier, your model predicts those who cancel their subscription with 99% accuracy and predicts those who renew their subscription with 82% accuracy. How should you interpret these results?
#162
You developed a Transformer model in TensorFlow to translate text. Your training data includes millions of documents in a Cloud Storage bucket. You plan to use distributed training to reduce training time. You need to configure the training job while minimizing the effort required to modify code and to manage the cluster’s configuration. What should you do?
#163
You work at a large organization that recently decided to move their ML and data workloads to Google Cloud. The data engineering team has exported the structured data to a Cloud Storage bucket in Avro format. You need to propose a workflow that performs analytics, creates features, and hosts the features that your ML models use for online prediction. How should you configure the pipeline?
#164
You are employed as an ML engineer at a social media company, and your current project involves creating a visual filter for users' profile photos. This entails training an ML model to identify bounding boxes around human faces. Your goal is to integrate this filter into your company's iOS-based mobile application with minimal code development while ensuring that the model is optimized for efficient inference on mobile devices. What steps should you take?
#165
While performing an exploratory analysis of a dataset, you come across a categorical feature A that exhibits significant predictive power. However, you notice that this feature is sometimes missing values. What course of action should you take?
#166
Your team frequently creates new ML models and runs experiments. Your team pushes code to a single repository hosted on Cloud Source Repositories. You want to create a continuous integration pipeline that automatically retrains the models whenever there is any modification of the code. What should be your first step to set up the CI pipeline?
#167
Your team has been assigned the responsibility of developing a machine learning (ML) solution within Google Cloud to categorize support requests for one of your platforms. After analyzing the requirements, you've chosen to utilize TensorFlow for constructing the classifier, providing you with complete control over the model's code, serving, and deployment. Your intention is to leverage Kubeflow pipelines as the ML platform to expedite the process. In order to save time, you aim to build upon existing resources and utilize managed services, rather than starting from scratch. How should you proceed in building the classifier?
#168
You're training and deploying updated versions of a regression model with tabular data using Vertex AI Pipelines, Vertex AI Training, Vertex AI Experiments, and Vertex AI Endpoints. The deployed model resides in a Vertex AI endpoint, and users access it via this endpoint. You aim to receive an email notification when significant changes occur in the feature data distribution, prompting you to retrigger the training pipeline and deploy an updated model. What should you do?
#169
You need to create an architecture for serving asynchronous predictions to detect potential failures in a mission-critical machine part. Your system collects data from various sensors on the machine. The goal is to build a model that can predict a failure occurring within the next N minutes based on the average sensor data over the past 12 hours. How should you go about designing this architecture?
#170
You are in the midst of training an ML model on a sizable dataset, and you are utilizing a TPU (Tensor Processing Unit) to accelerate the training process. However, you've noticed that the training is proceeding slower than expected, and upon investigation, you've determined that the TPU is not fully utilizing its capacity. What actions should you take to address this issue?
#171
You have recently developed a proof-of-concept (POC) deep learning model, and while you are satisfied with the overall architecture, you need to fine-tune a couple of hyperparameters. Specifically, you want to perform hyperparameter tuning on Vertex AI to determine the optimal values for the embedding dimension of a categorical feature and the learning rate. Here are the configurations you have set: - For the embedding dimension, you have defined the type as INTEGER with a range from a minimum value of 16 to a maximum value of 64. - For the learning rate, you have defined the type as DOUBLE with a range from a minimum value of 10e-05 to a maximum value of 10e-02. You are utilizing the default Bayesian optimization tuning algorithm, and your primary goal is to maximize the accuracy of the model. Training time is not a significant concern. In this context, how should you configure the hyperparameter scaling for each hyperparameter, and what should be the setting for maxParallelTrials?
#172
You are in the process of developing an MLOps platform to automate your company's machine learning experiments and model retraining. You require an efficient way to manage the artifacts for multiple pipelines. How should you go about organizing the pipelines' artifacts?
#1732টি বাছুন
You are in the process of building a model aimed at identifying fraudulent credit card transactions, with a primary focus on enhancing detection capabilities since overlooking even a single fraudulent transaction could have serious consequences for the credit card holder. To train this model, you have employed Vertex AI AutoML, utilizing users' profile details and credit card transaction data. However, after the initial model training, you've observed that the model is falling short in detecting a significant number of fraudulent transactions. What modifications should you make to the training parameters in Vertex AI AutoML to enhance the model's performance? (Select two options.)
#174
As an ML engineer at a worldwide shoe retailer, overseeing the company's website's machine learning models, you've been tasked with creating a recommendation model. This model should suggest new products to customers, taking into account their purchasing habits and similarities with other users. How should you proceed to build this model?
#175
You have recently employed XGBoost to train a Python-based model designed for online serving. Your model prediction service will be accessed by a backend service built in Golang, operating on a Google Kubernetes Engine (GKE) cluster. Your model necessitates both pre-processing and post-processing steps, which must be executed during serving. Your primary objectives are to minimize code alterations, reduce infrastructure maintenance, and expedite the deployment of your model into a production environment. What steps should you take to accomplish these goals?
#176
You possess a substantial collection of written support cases that fall into three distinct categories: Technical Support, Billing Support, or Other Issues. The task at hand is to efficiently construct, evaluate, and implement a system capable of automatically categorizing upcoming written requests into one of these predefined categories. How should you structure the pipeline to achieve this?
#177
You're employed by a magazine distributor and have the task of constructing a predictive model to forecast which customers will renew their subscriptions for the next year. You've utilized historical data from your organization as your training dataset and developed a TensorFlow model that has been deployed on AI Platform. Now, you must identify the customer attribute that holds the greatest predictive influence for each prediction generated by the model. How should you proceed with this task?
#178
You are tasked with building an MLOps pipeline to retrain tree-based models in production. The pipeline will include components related to data ingestion, data processing, model training, model evaluation, and model deployment. Your organization primarily uses PySpark-based workloads for data preprocessing. You want to minimize infrastructure management effort. How should you set up the pipeline?
#179
You are designing an ML pipeline for data processing, model training, and deployment using various Google Cloud services, and you have developed code for each task. Given the high frequency of new file arrivals, you need an orchestration layer that initiates only when new files appear in your Cloud Storage bucket. Additionally, you want to minimize compute node costs. What approach should you take?
#180
You hold the position of a senior ML engineer at a retail firm. Your objective is to establish a centralized method for monitoring and handling ML metadata, allowing your team to conduct reproducible experiments and generate artifacts. Which management solution should you suggest to your team?
