MLS-C01 Mock Exam 74 — 25 Questions

Question 1

A machine learning engineer is using Amazon SageMaker to train a deep learning model for image segmentation using the built-in Semantic Segmentation algorithm. The training dataset consists of 10,000 images with pixel-level annotations. The engineer notices that the model's validation loss is much higher than the training loss after 50 epochs, and the model performs poorly on a held-out test set. Which SageMaker training configuration change is MOST likely to improve the model's generalization?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Add data augmentation to the training pipeline, such as random cropping and flipping.. The model is overfitting, as indicated by high validation loss compared to training loss. Data augmentation is a powerful regularization technique that expands the effective training set with label-preserving transformations. For image segmentation, random cropping and flipping help the model generalize to variations in object position and orientation. Early stopping only stops training but doesn't improve the model's ability to generalize; increasing epochs or reducing learning rate do not address the root cause.

Answer

Reduce the learning rate to make the training more stable.

Answer

Increase the number of epochs to 100 to allow the model to learn more from the training data.

Answer

Enable early stopping with a patience parameter based on validation loss.

Question 2

A data scientist notices that one numerical feature in a SageMaker training dataset is heavily right-skewed and contains a small number of extreme high values. Before modeling, the scientist wants to reduce the influence of those extremes while preserving relative ordering of observations. Which transformation should the scientist apply to this feature?

Accepted Answer

Apply a logarithmic transformation to the feature values.. The scenario requires reducing the influence of extreme values in a right-skewed continuous feature while keeping relative order. A logarithmic transformation is monotonic and compresses the upper tail most strongly, which directly lessens the leverage of large observations without changing how observations rank against each other. Scaling approaches only change units, and encoding approaches are meant for categorical data, so neither addresses skew or outlier influence.

Answer

Apply one-hot encoding to split the feature into binary indicator columns.

Answer

Apply min-max scaling to rescale the feature into the range zero to one.

Answer

Apply z-score standardization to center the feature at zero with unit variance.

Question 3

A data scientist must explore a 25 GB dataset stored in Amazon S3 to compute summary statistics and correlation coefficients before modeling. The data is already cataloged in the AWS Glue Data Catalog. The scientist wants to write standard SQL and avoid loading the full dataset into memory. Which approach is MOST appropriate?

Accepted Answer

Run SQL queries with Amazon Athena against the cataloged table and inspect the results. Athena is serverless and reads S3 data in place through the Glue Data Catalog, letting the scientist write SQL aggregates including corr for correlation without moving or fully materializing the data. Local pandas loading, Redshift COPY, and Kinesis Data Analytics all either exceed the memory constraint, duplicate the data, or target the wrong data paradigm.

Answer

Download all objects to a local workstation and load them into pandas DataFrames

Answer

Use Amazon Kinesis Data Analytics to run SQL over the S3 objects

Answer

Create an Amazon Redshift cluster and COPY the S3 data into it before querying

Question 4

A data scientist is performing exploratory data analysis on a dataset with a numerical target variable and several numerical features. The scientist notices that one feature has a highly skewed distribution and decides to apply a logarithmic transformation to reduce skewness. After the transformation, the scientist wants to assess the linear relationship between the transformed feature and the target variable. Which statistical measure should the scientist use to evaluate the strength and direction of the linear relationship?

Accepted Answer

Pearson correlation coefficient.. Pearson correlation coefficient is the correct choice because it quantifies the strength and direction of a linear relationship between two continuous variables. After a logarithmic transformation to reduce skewness, the relationship may become more linear, and Pearson correlation is designed to measure that. The other measures are either rank-based or capture non-linear associations, which do not directly address the linearity goal.

Answer

Kendall's tau correlation coefficient.

Answer

Spearman rank correlation coefficient.

Answer

Distance correlation.

Question 5

A machine learning team stores 40 TB of historical training features in an Amazon S3 bucket as uncompressed CSV. A SageMaker training job reads the full dataset at the start of every run using File mode, and each run now spends over two hours just downloading data before training begins. The team wants to reduce the startup time of each training job with minimal changes to the training script. What should a data engineer do?

Accepted Answer

Convert the CSV objects to Apache Parquet and use Pipe mode with the SageMaker training job.. Pipe mode streams records from S3 on demand rather than copying the whole dataset to local storage before training, eliminating the large upfront download characteristic of File mode. Pairing it with Parquet reduces the bytes transferred because only needed columns are read. Together these changes cut startup time substantially while requiring at most a small change to the input configuration, not a rewrite of the training logic.

Answer

Move the dataset into an Amazon EFS file system mounted on the training instances and keep File mode.

Answer

Increase the SageMaker training instance's attached Amazon EBS volume size and keep File mode.

Answer

Enable S3 Transfer Acceleration on the bucket and keep File mode with CSV input.

Question 6

A data scientist is training a deep learning model for image segmentation using Amazon SageMaker. The model is a fully convolutional network (FCN) with a ResNet-50 backbone. The training dataset contains 10,000 images with pixel-level annotations. The data scientist notices that the model performs well on the training set but poorly on the validation set. Which technique should the data scientist apply to reduce overfitting?

Accepted Answer

Add dropout layers after each convolutional block and apply L2 regularization.. Overfitting occurs when a model memorizes training data but fails to generalize. For a deep convolutional network like FCN with ResNet-50, adding dropout layers and L2 regularization are standard techniques to reduce overfitting. Dropout prevents co-adaptation of neurons, while L2 penalizes large weights. Together, they constrain the model's capacity and improve validation performance without sacrificing training accuracy excessively.

Answer

Increase the learning rate and use a larger batch size.

Answer

Reduce the image resolution and use data augmentation.

Answer

Use batch normalization and increase the number of epochs.

Question 7

A data scientist has a cleaned dataset stored as a single CSV in Amazon S3 and wants quick, code-free summary statistics, distributions, and correlation insights before starting feature engineering. The scientist prefers a visual interface inside SageMaker Studio over writing pandas code. Which SageMaker capability should be used?

Accepted Answer

Amazon SageMaker Data Wrangler data quality and insights report. The data quality and insights report in Data Wrangler is purpose-built for visual, code-free profiling of a dataset, producing summary statistics, distributions, missing-value analysis, and anomaly or leakage detection. Clarify, Model Monitor, and Experiments all operate around models, endpoints, or training runs, so none of them profiles a standalone CSV in the way this scenario demands.

Answer

Amazon SageMaker Clarify explainability report

Answer

Amazon SageMaker Experiments trial component charts

Answer

Amazon SageMaker Model Monitor data quality monitoring schedule

Question 8

A machine learning engineer is exploring a dataset with a highly skewed numerical feature 'income'. The feature has a long right tail and several extreme outliers. She wants to apply a transformation to make the distribution more symmetric before feeding it into a linear regression model. Which transformation should she apply using Amazon SageMaker Data Wrangler?

Accepted Answer

Log transformation. For a right-skewed feature with outliers, a log transformation is effective at reducing skewness and stabilizing variance. It compresses large values and expands smaller ones, making the distribution more symmetric. This helps linear models that assume normally distributed errors. Scaling methods like standard, min-max, or robust scaling only change the scale, not the shape, so they do not address skewness.

Answer

Standard scaler

Answer

Robust scaler

Answer

Min-max scaler

Question 9

A data scientist is training a deep learning model using Amazon SageMaker. The training data is stored in Amazon S3 and is approximately 2 TB in size. The training job is taking a long time to start because the data is being downloaded from S3 to the training instance's Amazon EBS volume. The data scientist wants to reduce the startup time and improve training performance. Which two actions should the data scientist take? (Choose two.)

Accepted Answer

Use Amazon FSx for Lustre as a data source and mount it to the training instance.. Using Pipe mode streams data directly from S3, eliminating the initial download delay, while Amazon FSx for Lustre provides a high-performance file system that allows immediate access to data without full download. Both actions significantly reduce startup time and improve training performance for large datasets. They are complementary and can be used together for optimal results.

Answer

Enable Amazon SageMaker Managed Spot Training to reduce cost.

Answer

Compress the training data into a single .tar.gz file and use File mode.

Answer

Increase the size of the Amazon EBS volume attached to the training instance.

Question 10

A machine learning engineer has trained a SageMaker model and registered it in the Amazon SageMaker Model Registry. Compliance requires that every production endpoint only serve model versions that have been reviewed and approved, and that no unreviewed artifact can ever be deployed to production. The engineer needs to wire the deployment pipeline so that only approved model versions are deployed. Which approach meets these requirements with the LEAST operational effort?

Accepted Answer

Use the SageMaker Model Registry model package group's approval status; configure the deployment pipeline to deploy only model package versions whose ModelApprovalStatus is Approved.. The SageMaker Model Registry is the designed control point for model governance: each registered model package version carries a ModelApprovalStatus that a CI/CD pipeline can query before deployment. Deploying strictly versions with Approved status creates an auditable, low-maintenance gate that satisfies the compliance requirement without inventing custom approval tracking or file-move rituals.

Answer

Enable SageMaker Model Monitor on the endpoint and configure a baseline that alerts when an unapproved model is detected serving traffic.

Answer

Copy each candidate model artifact to a separate S3 prefix named approved/ only after a reviewer manually re-uploads the tarball, and have the pipeline deploy from that prefix.

Answer

Attach an IAM policy to the endpoint's execution role that denies sagemaker:CreateModel unless the model name contains the string approved.

Question 11

A fraud-detection team at a bank trains an XGBoost binary classifier on 2 million transactions where only 0.8% are fraudulent. The model achieves 0.995 AUC but the fraud team reports that at the default 0.5 probability threshold, recall for the fraud class is 0.31, meaning most frauds are missed. The team wants to increase fraud recall without retraining the model. Which approach should they take?

Accepted Answer

Adjust the classification threshold used at inference time below 0.5 to trade precision for higher recall on the minority class.. When a classifier ranks well but operates at an inappropriate cut point, the operating point is the problem, not the model. For rare-event detection, the default 0.5 threshold is rarely aligned with the business cost of false negatives. Adjusting the threshold at inference time raises minority-class recall using the existing model, satisfying the constraint of no retraining.

Answer

Increase the number of boosting rounds until the validation AUC exceeds 0.999, then redeploy the model to the endpoint.

Answer

Retrain the model with the scale_pos_weight hyperparameter set to a value near 1.0 so the two classes are weighted equally.

Answer

Apply SMOTE to the training data and retrain so the positive class represents roughly 50% of the samples.

Question 12

A machine learning specialist is using Amazon SageMaker to train a model with the built-in k-means algorithm. The dataset contains numerical features with varying scales, and the specialist wants to ensure that the clustering is not dominated by features with larger magnitudes. Which preprocessing step should the specialist perform before training?

Accepted Answer

Normalize the features to have zero mean and unit variance.. k-means relies on distance calculations, so features with larger scales can dominate the clustering. Normalizing features to zero mean and unit variance ensures equal contribution from all features, leading to more meaningful clusters. This is a standard and necessary preprocessing step when features have different units or ranges.

Answer

Convert all features to categorical bins.

Answer

Apply one-hot encoding to all features.

Answer

Apply principal component analysis (PCA) to reduce dimensionality.

Question 13

A data scientist is using Amazon SageMaker to train a deep learning model with the built-in Object Detection algorithm. The training dataset contains images with bounding boxes for multiple object classes. The scientist notices that the model performs well on the training set but poorly on the validation set, and the validation loss increases after a few epochs. Which action should the scientist take to improve generalization?

Accepted Answer

Enable data augmentation and increase the weight decay hyperparameter.. Data augmentation increases the diversity of training data without collecting new images, helping the model generalize better. Increasing weight decay applies stronger L2 regularization, which penalizes complex models. Both are standard techniques to combat overfitting, as indicated by the rising validation loss and large train-validation gap.

Answer

Reduce the number of training epochs and increase the number of classes.

Answer

Use a larger pre-trained model and freeze all layers.

Answer

Increase the learning rate and reduce the batch size.

Question 14

A data scientist is exploring a dataset in Amazon SageMaker Studio and wants to quickly generate summary statistics and visualizations for all features, including histograms and scatterplots, without writing code. Which SageMaker Studio feature should they use?

Accepted Answer

SageMaker Data Wrangler. SageMaker Data Wrangler is a visual data preparation and EDA tool within SageMaker Studio. It allows users to import data, generate summary statistics, and create visualizations like histograms and scatterplots without writing code. The other SageMaker features serve different purposes: Experiments for tracking, Debugger for training diagnostics, and Autopilot for automated model building.

Answer

SageMaker Debugger

Answer

SageMaker Experiments

Answer

SageMaker Autopilot

Question 15

A company ingests streaming sensor data into Amazon Kinesis Data Streams and must store it in Amazon S3 in Parquet for analytics. The data volume varies significantly, with long idle periods overnight and bursts during the day. The solution must minimize cost during idle periods and scale automatically during bursts, with minimal operational management. Which solution meets these requirements?

Accepted Answer

Use Amazon Kinesis Data Firehose to deliver the stream to S3 with a Lambda function for Parquet conversion, and enable dynamic partitioning.. Kinesis Data Firehose is a fully managed delivery service that scales automatically with incoming throughput and bills by data volume, so idle overnight periods incur minimal cost. Lambda provides the Parquet conversion, and dynamic partitioning organizes output into S3 prefixes for efficient querying. This meets the variable-load, cost, and low-management requirements without provisioning capacity.

Answer

Provision a fixed number of Kinesis Data Streams shards sized for peak load and run an AWS Lambda consumer that converts records to Parquet and writes to S3.

Answer

Use AWS Glue streaming ETL jobs with a fixed number of workers to read from Kinesis and write Parquet to S3.

Answer

Run a continuously provisioned Amazon EMR cluster with Spark Streaming that reads from Kinesis and writes Parquet to S3.

Question 16

A financial analytics team stores daily transaction records in an Amazon S3 bucket as Parquet files. They need to perform exploratory data analysis on a 2 TB dataset to compute summary statistics and visualize correlations between features. The team wants to minimize data movement and cost while using a familiar SQL-like interface. Which AWS service should they use?

Accepted Answer

Amazon Athena. Amazon Athena provides a serverless, SQL-based way to analyze data directly in Amazon S3, which is well-suited for exploratory data analysis on large Parquet datasets. It avoids data movement and infrastructure management, making it cost-effective for computing summary statistics and correlations.

Answer

AWS Glue DataBrew

Answer

Amazon SageMaker Data Wrangler

Answer

Amazon Redshift Spectrum

Question 17

A data scientist has a .csv file containing 500,000 rows of customer churn data stored in Amazon S3. Before building any model, she wants a quick, automated visual profile of every column — including histograms, quantile statistics, and a correlation matrix — without writing any code. Which solution should she use?

Accepted Answer

Create a SageMaker Data Wrangler flow, import the S3 data source, and use the built-in Data Quality and Insights report.. Amazon SageMaker Data Wrangler's Data Quality and Insights report is the only listed capability that ingests S3 data and automatically emits histograms, quantile statistics, and a correlation matrix with no coding. SageMaker Processing, Athena, and DynamoDB all require the analyst to build the analysis herself and do not natively produce that combined automated profile.

Answer

Open the file in Amazon Athena, run SELECT * and inspect the query results grid for distribution insights.

Answer

Load the file into Amazon DynamoDB and enable DynamoDB Streams to capture column-level statistics.

Answer

Run an Amazon SageMaker Processing job with scikit-learn to generate the plots and write them to S3.

Question 18

A data analyst is exploring a dataset containing information about customer orders. The dataset includes a column 'order_date' with timestamps and a column 'order_total' with numerical values. The analyst wants to visualize how the average order total changes over time to identify trends and seasonality. Which type of plot is most appropriate for this exploratory analysis?

Accepted Answer

A line plot with time on the x-axis and average order total on the y-axis.. A line plot with time on the x-axis and average order total on the y-axis is the most appropriate because it aggregates data over time and connects points to reveal trends and seasonality. It provides a clear visual of how the average changes continuously, making it easy to spot patterns. Other plots either do not aggregate, lack the time dimension, or treat time as discrete.

Answer

A box plot of order_total grouped by month.

Answer

A histogram of order_total values.

Answer

A scatter plot with order_date on the x-axis and order_total on the y-axis.

Question 19

A data scientist is tuning a SageMaker XGBoost model and wants to search over learning rate, max depth, and subsample simultaneously. The search space is large and each training job takes about 40 minutes. The scientist needs to minimize total tuning cost while still finding a strong configuration. Which SageMaker automatic model tuning strategy should be used?

Accepted Answer

Bayesian optimization. When each training job is expensive and the search space spans several interacting hyperparameters, a strategy that learns from completed trials is most cost-effective. Bayesian optimization fits a surrogate model to observed objective values and proposes promising candidates, reducing the number of full training runs needed. Random and grid search ignore trial outcomes, and Hyperband's early stopping is designed for cheaper, epoch-wise evaluations than the 40-minute jobs described.

Answer

Grid search

Answer

Hyperband

Answer

Random search

Question 20

A machine learning engineer needs to train a model on a 2 TB dataset stored in Amazon S3 as CSV files. The training job will run on an Amazon SageMaker training instance. The engineer wants to minimize training time and avoid downloading the entire dataset to the instance's Amazon EBS volume. Which data input mode should be used?

Accepted Answer

FastFile mode. FastFile mode allows SageMaker training jobs to stream data from S3 as if it were a local file system, avoiding the need to download the full dataset to EBS. This reduces startup time and disk requirements, which is ideal for large datasets. Pipe mode also streams but has stricter algorithm compatibility requirements. FastFile offers a balance of performance and compatibility.

Answer

Augmented manifest file mode

Answer

Pipe mode

Answer

File mode