Timed Mode Diagnostic Test – AWS Certified Machine Learning

A Machine Learning Specialist has a custom algorithm bundled in a TensorFlow Docker container. The Specialist needs to configure the container so that it could be run as executable by Amazon SageMaker.

Which course of action should the Specialist take?

An electronics manufacturing company is building an image classification model to detect whether a circuit board is defective or not. To decrease training time, a Machine Learning Specialist implemented transfer learning using 300 images on a pre-trained neural network obtained from a partner company. The company requires a model accuracy of no less than 95% to be feasible. The Specialist conducted a full grid search through a wide hyperparameter space. However, the optimal values returned only 73% and 72% accuracy on the training and validation set, respectively.

How can the Specialist increase the model accuracy?

A Machine Learning Specialist is developing a linear regression model to predict the average salary of ML-engineers in Asia. The Specialist will use a p-value of 0.05 to test the statistical significance of each predictor within a dataset. A quick visualization of the data shows that nearly all predictors are distributed normally. A column in the dataset is plotted below:

Which transformation should the Specialist apply to the column to meet the model’s statistical assumption?

A web service is producing thousands of records into an Amazon S3 bucket every minute. A Data Engineering team performs queries against these data using Amazon Athena. Due to the large volume of data, most of the SQL queries run slowly. The team seeks a solution that will improve query performance.

How should the solution be implemented?

A company is planning an employee retention strategy to reduce replacement costs. The company hires a Machine Learning Specialist to develop an ML-based solution to tell whether a current employee will leave or stay for the next 6 months. The ML Specialist was given full access to labeled datasets.

How should the specialist accomplish this task?

A footwear company is using an Amazon Personalize-powered solution that gives tailored recommendations and content to its customers based on purchase and browsing history. Initially, the company observes a steady increase in sales of recommended products to current customers. However, after some time, the sales have started to decline, which prompted the company to deploy a new solution version.

How should the company configure the new solution?

A retail company is storing structured and semi-structured data in an S3 bucket. A Machine Learning Specialist is required to analyze the data using SQL queries. The company needs a solution with the least amount of operational overhead.

Which approach satisfies the requirement?

A company is running deep neural networks on GPU-based instances for their product recommendation system. The system relies on an ML-workflow that fetches each dataset from an S3 bucket, loads it into a Keras model pulled from a local repository, then continuously delivers the results into a separate bucket. The current workflow is expensive to run and involves manual resource management.

The ML-team that supervises the system wants to have the least cost and operating overhead. They’re looking for an automated solution with the following requirements:

• Given that a job requires several hours of on-going work, in case of a failure, a job should be able to retry automatically.

• Each job must be submitted to a central queue.

• The workflow will only run once a week, every Monday to Friday.

How should the ML-team redesign the workflow?

A Machine Learning Specialist is wondering if he should build a full Bayesian Network or a Naive Bayesian model for a classification task. The computed Pearson correlation coefficients between each predictor range from 0.1 – 0.97.

Based on this data, which model should the Specialist choose and why? (Select TWO.)

A Data Engineer is implementing a critical long-running task in an Amazon EMR cluster. After estimating its workload, the Engineer has decided to configure the cluster with a single master node, 5 core nodes, and 10 task nodes. He plans to use Spot instances to further reduce costs.

Which node should the Engineer choose to launch the Spot instances?

A Data Engineer needs to build a data repository to store hundreds of existing training datasets for model training. The repository should be available to all members of a Machine Learning team for storing new datasets. The storage scheme should automatically scale and have to be cost-efficient. It should also support SQL querying capabilities while ensuring secure access and governance over the data.

Which method of storing data is the MOST suitable?

A Machine Learning Specialist has created a model without performing an initial investigation on the training data. The training time is slow and the predictive ability of the model is unstable during inference. A co-worker recommended that he should run the Principal Component Analysis (PCA) algorithm on the data before starting to train the model.

Which is a valid effect of the said algorithm?

A Machine Learning Specialist has completed training on a deep learning model built with Apache MXNet on Amazon SageMaker AI. The Specialist wants to compile and deploy this model to an ARM-based Android phone for inference.

Which service should the Specialist use to accomplish its goal?

A company wants to add a data ingestion system to an existing machine learning pipeline using Amazon Kinesis Data Streams. The data stream is expected to collect 8 KB of JSON data at up to 1,000 transactions per second.

What should be the minimum number of shards to be configured to support the ingestion rate?

A Machine Learning Specialist is planning to train a model using Amazon SageMaker to determine the impact of weather conditions on the air quality of a city. The ML Specialist wants to forecast the NO₂ concentration for the next 7 days. She will be using 5 years’ worth of data that has roughly about 40,000 observations. As much as possible, the model must provide cheap inference costs.

Which algorithm should be used to build the model?

A Machine Learning Specialist is preparing a sandbox environment for the Data Analytics team to access notebooks and train models in Amazon SageMaker. The Administrator should be able to track and record the number of deployments performed, the CPU and GPU utilization of the SageMaker instances, as well as the number of invocation errors.

Which combination of services could provide this information? (Select TWO.)

A car insurance company seeks to automate car damage assessment in claims processing using machine learning. The company plans to build a deep learning model for image analysis from a pre-trained Convolutional Neural Network (CNN) model using transfer learning. To improve accuracy, the company will fine-tune the model against a custom dataset using Keras via the Amazon SageMaker SDK.

The company wants to optimize prediction times by reducing the model size without compromising accuracy.

How can the company fulfill the requirements in the most efficient way?

A data scientist is preparing the training data for a regression model that will estimate the resale value of a used car. The data contains the following set of key features:

• Resale price ($1,500 – $50,000)

• Build year (2005 – 2021)

• Mileage (100 Km – 200,000 Km)

• Transmission (Automatic/Manual)

• Fuel Type (Petrol/Diesel)

• Engine Size (1.3L – 2.5L)

The dataset follows a nearly normal distribution and has few outliers.

Which combination of methods is the most appropriate way of preparing the data effectively?

A machine learning (ML) engineer has deployed an AI service using an Amazon SageMaker instance. The model’s endpoint receives varying amounts of traffic throughout the day. Some users have reported cases where the service becomes inaccessible.

The ML engineer decides to implement auto-scaling for the SageMaker endpoint to ensure the service can keep up with the demand.

Which metric should the ML engineer use for tracking?

A data scientist is building a linear regression model to predict the performance of fantasy basketball picks. The scientist is using a correlation matrix to examine the relationship between various player stats and fantasy points.

The correlation matrix is shown below:

Which statements are true about the matrix? (Select Two)

A company has a team of data analysts, including both technical and non-technical members. The company seeks an easier and faster way for the team to perform data preprocessing tasks, such as normalization and filling in missing values. The company prefers to minimize overhead and avoid managing infrastructure.

Which solution meets the requirement?