View Full Microsoft AI-901 Exam Dumps and Practice Test Dumps

 

Question 121. Which Azure service provides capabilities for building and managing machine learning models?

1. Azure AI Speech
2. Azure AI Translator
3. Azure Machine Learning
4. Azure AI Vision

Correct Answer: 3. Azure Machine Learning

Explanation:

Azure Machine Learning is a cloud-based service designed to support the development, training, evaluation, deployment, and management of machine learning models. Data scientists and developers can use it to create experiments, train models, manage datasets and models, deploy models to endpoints, and monitor machine learning workloads. Azure AI Speech, Azure AI Translator, and Azure AI Vision are specialized AI services focused on speech, language translation, and visual analysis respectively. Azure Machine Learning is more appropriate when an organization needs to build or operate its own machine learning workflows. Understanding this distinction is important when selecting an Azure service for AI-901 scenarios.

Question 122. Which type of learning uses labeled examples to train a model?

1. Unsupervised learning
2. Supervised learning
3. Reinforcement learning
4. Clustering

Correct Answer: 2. Supervised learning

Explanation:

Supervised learning uses labeled training examples to teach a machine learning model how to make predictions. Each training example generally contains input features along with a known target or label. The model learns relationships between the inputs and expected outputs and can then make predictions for new data. Classification and regression are common supervised learning tasks. Unsupervised learning works with data that does not have predefined target labels, while clustering is one example of an unsupervised technique. Reinforcement learning uses feedback such as rewards or penalties to guide behavior. Therefore, labeled training examples are a defining characteristic of supervised learning.

Question 123. Which type of machine learning discovers patterns in data without predefined labels?

1. Supervised learning
2. Reinforcement learning
3. Unsupervised learning
4. Classification

Correct Answer: 3. Unsupervised learning

Explanation:

Unsupervised learning analyzes data without requiring predefined target labels. Instead of learning to predict a known outcome, the algorithm attempts to discover useful structures, relationships, or patterns within the available data. Clustering is a common example because it groups similar records based on their characteristics without requiring predefined categories. Supervised learning requires labeled examples and is commonly used for classification and regression. Reinforcement learning is based on interactions with an environment and feedback. Therefore, if a scenario describes discovering natural patterns or groups in unlabeled data, unsupervised learning is the appropriate machine learning approach.

Question 124. Which learning approach uses rewards or penalties to guide an agent’s behavior?

1. Reinforcement learning
2. Regression
3. Classification
4. Unsupervised learning

Correct Answer: 1. Reinforcement learning

Explanation:

Reinforcement learning trains an agent through interaction with an environment. The agent takes actions and receives feedback in the form of rewards or penalties. Over time, it learns a strategy intended to maximize the accumulated reward. This approach differs from supervised learning, where models learn from labeled examples. It also differs from unsupervised learning, where the objective is generally to discover patterns in unlabeled data. Reinforcement learning can be applied to areas such as game playing, robotics, and decision-making systems. For AI-901, a scenario involving an agent learning through rewards and penalties is describing reinforcement learning.

Question 125. What is a feature in a machine learning dataset?

1. The final prediction made by a model
2. An input variable used by a model
3. A model deployment endpoint
4. A machine learning algorithm

Correct Answer: 2. An input variable used by a model

Explanation:

A feature is an input variable that provides information to a machine learning model. Features are used by the model to learn patterns and generate predictions. For example, a house-price model might use features such as property size, number of bedrooms, location, and age of the property. The target or label represents the value the model is intended to predict in supervised learning. Features should therefore be distinguished from predictions, algorithms, and deployment endpoints. Understanding features is fundamental to machine learning because the quality and relevance of input features can influence how effectively a model learns patterns and produces useful predictions.

Question 126. In supervised learning, what does the label represent?

1. The input feature name
2. The amount of compute available
3. The target value the model learns to predict
4. The model endpoint URL

Correct Answer: 3. The target value the model learns to predict

Explanation:

In supervised learning, a label represents the known target value associated with a training example. The machine learning model uses labeled examples to learn a relationship between input features and the target. For a classification problem, the label could be a category such as “spam” or “not spam.” For regression, the label could be a numerical value such as a house price. Features provide the input information, while the label represents the expected output. Compute resources and endpoint addresses are infrastructure concepts rather than labels. Therefore, the target value that the model is trained to predict is the label.

Question 127. What is overfitting in machine learning?

1. When a model performs well on training data but poorly on unseen data
2. When a model has no training data
3. When all predictions are identical
4. When a dataset contains only numerical values

Correct Answer: 1. When a model performs well on training data but poorly on unseen data

Explanation:

Overfitting occurs when a machine learning model learns the training data too closely, including patterns or noise that do not generalize to new examples. As a result, the model may perform very well on its training dataset but show significantly poorer performance on validation or test data. Overfitting can reduce the usefulness of a model in real-world scenarios because real inputs are usually different from the exact examples used during training. Techniques such as using more representative data, regularization, cross-validation, feature selection, or adjusting model complexity can help address overfitting. For AI-901, strong training performance combined with weak unseen-data performance is a key indicator of overfitting.

Question 128. What is underfitting in machine learning?

1. When a model is too complex for the available hardware
2. When a model memorizes every training example
3. When a model performs poorly because it has not learned the underlying patterns sufficiently
4. When a model is deployed successfully

Correct Answer: 3. When a model performs poorly because it has not learned the underlying patterns sufficiently

Explanation:

Underfitting occurs when a machine learning model is too simple or insufficiently trained to capture important patterns in the data. In this situation, the model may perform poorly not only on unseen data but also on the training dataset itself. Underfitting can result from an overly simple model, insufficiently informative features, excessive regularization, or inadequate training. This is different from overfitting, where the model learns the training data too closely and then struggles with new examples. For AI-901, if both training and validation performance are poor because the model has failed to capture important relationships, underfitting is a likely explanation.

Question 129. Which dataset is primarily used to train the parameters of a machine learning model?

1. Test dataset
2. Validation dataset
3. Production dataset
4. Training dataset

Correct Answer: 4. Training dataset

Explanation:

The training dataset is used by a machine learning algorithm to learn model parameters from examples. During training, the model examines the input features and, in supervised learning, the associated labels to identify patterns that can be used to make predictions. Validation data can be used during development to compare models or tune settings, while test data is generally reserved for final evaluation on unseen examples. Production data is generated or encountered after deployment and is not normally the primary dataset used to train the initial model. Therefore, when a question asks which dataset is used to learn the model’s parameters, the answer is the training dataset.

Question 130. Which dataset can be used during model development to tune hyperparameters and compare model configurations?

1. Validation dataset
2. Test dataset
3. Production dataset
4. Archived dataset

Correct Answer: 1. Validation dataset

Explanation:

A validation dataset is commonly used during model development to evaluate different model configurations and help tune hyperparameters. Developers can train candidate models using the training data and then compare their performance on validation data. This provides information that can help select an appropriate configuration without using the final test dataset for repeated development decisions. The test dataset should generally be kept separate so that it can provide a more unbiased estimate of performance after development is complete. Production data serves operational purposes and is not automatically a validation set. Therefore, validation data is the appropriate choice for tuning and comparing model configurations.

Question 131. Which metric measures the proportion of all predictions that are correct?

1. Recall
2. Precision
3. Accuracy
4. MAE

Correct Answer: 3. Accuracy

Explanation:

Accuracy measures the proportion of predictions that a classification model gets correct out of all evaluated predictions. It can be calculated as the number of correct predictions divided by the total number of predictions. Accuracy can be useful when the classes are reasonably balanced and the costs of different types of errors are similar. However, accuracy alone may not provide enough information when classes are highly imbalanced or when false positives and false negatives have different consequences. Precision focuses on predicted positives, while recall focuses on actual positives. MAE is commonly used for regression. Therefore, the overall proportion of correct classification predictions is measured by accuracy.

Question 132. Which metric is commonly used to measure the average squared difference between predicted and actual numerical values?

1. Recall
2. Mean squared error
3. Precision
4. Accuracy

Correct Answer: 2. Mean squared error

Explanation:

Mean squared error, or MSE, is a regression metric that calculates the average of the squared differences between predicted values and actual values. Squaring the errors makes larger errors contribute disproportionately to the final value, which can make MSE useful when larger mistakes should receive greater emphasis. Because the errors are squared, the resulting metric is expressed in squared units of the target variable. MAE instead uses absolute differences and is often easier to interpret in the original units. Precision, recall, and accuracy are classification metrics. Therefore, when a question asks for the average squared prediction error in a regression problem, mean squared error is the appropriate metric.

Question 133. Which Azure Machine Learning component can execute machine learning jobs using scalable cloud compute?

1. Compute cluster
2. Azure AI Translator
3. Azure AI Vision
4. Azure AI Language

Correct Answer: 1. Compute cluster

Explanation:

An Azure Machine Learning compute cluster provides scalable cloud-based processing resources for machine learning workloads. It can be used to run training jobs and other compute-intensive tasks without requiring developers to maintain physical machines. Compute resources can scale according to workload requirements, helping teams handle experiments or training processes that require more processing capacity. Azure AI Translator, Azure AI Vision, and Azure AI Language are specialized AI services and are not Azure Machine Learning compute resources. Therefore, when an AI-901 scenario describes running machine learning jobs using scalable cloud processing, an Azure Machine Learning compute cluster is a suitable choice.

Question 134. Why should machine learning data be representative of the population on which the model will be used?

1. To guarantee zero prediction errors
2. To reduce the need for any evaluation
3. To ensure the model can be deployed without testing
4. To help the model learn patterns that generalize to intended users

Correct Answer: 4. To help the model learn patterns that generalize to intended users

Explanation:

Representative training data helps a machine learning model learn patterns that reflect the population and situations in which the model will eventually operate. If important groups or scenarios are missing or severely underrepresented, the model may perform differently across the real-world population than expected from its development results. Representative data does not guarantee perfect predictions or eliminate the need for evaluation, but it can improve the relevance and reliability of the learned patterns. Data quality, diversity, labeling, and coverage should all be considered when preparing training datasets. Therefore, representative data is important because it supports better generalization to the intended users and operating conditions.

Question 135. Which responsible AI principle focuses on ensuring AI systems produce equitable outcomes across different groups?

1. Fairness
2. Transparency
3. Accountability
4. Reliability

Correct Answer: 1. Fairness

Explanation:

Fairness in responsible AI focuses on ensuring that AI systems do not produce unjustified or discriminatory differences in outcomes for different groups. Developers should consider whether training data, features, labels, and model behavior could contribute to unequal performance or treatment. Evaluating model performance across relevant groups can help identify potential disparities. Fairness does not mean that every group must always receive identical outcomes, because legitimate differences in circumstances can exist. Instead, the goal is to avoid unjustified bias and unequal treatment. Transparency focuses on understanding system behavior, accountability concerns responsibility, and reliability concerns dependable operation. Therefore, fairness is the principle most directly related to equitable outcomes.

Question 136. Which responsible AI principle is concerned with making AI systems dependable and resistant to unexpected failures?

1. Inclusiveness
2. Fairness
3. Reliability and safety
4. Transparency

Correct Answer: 3. Reliability and safety

Explanation:

Reliability and safety concern whether an AI system can operate consistently and safely under expected conditions and handle unexpected situations appropriately. AI systems may encounter unusual inputs, changing environments, or circumstances that were not represented fully in training data. Developers should therefore evaluate performance, test systems under relevant conditions, monitor deployed solutions, and implement safeguards where appropriate. Fairness addresses equitable outcomes, inclusiveness focuses on serving people with different needs, and transparency concerns communicating how systems work and what their limitations are. For AI-901, a requirement involving dependable behavior, safe operation, robustness, and handling unexpected conditions relates directly to reliability and safety.

Question 137. What is the purpose of monitoring a deployed machine learning model?

1. To permanently change all training labels
2. To detect performance, data, or operational issues
3. To eliminate the need for testing
4. To convert predictions into images

Correct Answer: 2. To detect performance, data, or operational issues

Explanation:

Monitoring a deployed machine learning model helps organizations identify issues that may occur after deployment. These issues can include changes in incoming data, degraded model performance, operational failures, unusual request patterns, or other conditions that affect the solution. A model that performed well during development may encounter different data or circumstances in production, so ongoing monitoring is an important part of the model lifecycle. Monitoring does not replace testing and does not automatically change training labels. Instead, monitoring provides information that can help teams determine when investigation, retraining, configuration changes, or other actions may be required. Therefore, detecting performance, data, and operational problems is a primary purpose of monitoring.

Question 138. What is model drift?

1. A situation where the model’s performance or data relationship changes over time
2. A method for converting speech into text
3. A technique for creating image captions
4. A process for deleting unused models

Correct Answer: 1. A situation where the model’s performance or data relationship changes over time

Explanation:

Model drift refers to changes that can cause a machine learning model’s effectiveness to decrease over time. Real-world data and relationships between inputs and outcomes can change after a model has been deployed. For example, customer behavior, market conditions, or other environmental factors may change, causing patterns learned during training to become less representative of current conditions. Monitoring can help identify these changes and determine whether a model needs investigation or retraining. Model drift is unrelated to speech transcription, image captioning, or simply deleting models. For AI-901, a decline in model effectiveness caused by changing real-world conditions is an important reason to monitor deployed machine learning solutions.

Question 139. Which Azure AI service can help detect potentially harmful or unsafe text content?

1. Azure AI Speech
2. Azure AI Translator
3. Azure AI Content Safety
4. Azure AI Vision

Correct Answer: 3. Azure AI Content Safety

Explanation:

Azure AI Content Safety provides capabilities for detecting and managing potentially harmful content in AI applications. It can help analyze text and other supported content for categories of potentially harmful material and can be incorporated into applications that need content moderation safeguards. This can be particularly useful when building applications where users can submit content or where generative AI systems produce responses that need to be evaluated. Azure AI Speech handles speech processing, Translator handles language translation, and Vision focuses on visual analysis. Therefore, when an AI-901 scenario requires an Azure service specifically intended to help identify potentially harmful content, Azure AI Content Safety is the relevant service.

Question 140. Which practice can help protect sensitive information when developing an AI solution?

1. Collect every available piece of personal data
2. Publish sensitive training records publicly
3. Disable access controls
4. Minimize collected data and apply appropriate access controls

Correct Answer: 4. Minimize collected data and apply appropriate access controls

Explanation:

Protecting sensitive information requires organizations to consider how data is collected, stored, accessed, processed, and retained throughout the AI solution lifecycle. Data minimization means collecting only the information necessary for the intended purpose rather than gathering unnecessary personal information. Appropriate access controls can then help ensure that only authorized users or services can access sensitive resources. Additional measures may include encryption, secure storage, monitoring, and suitable retention policies. Collecting excessive data or publishing sensitive records increases privacy and security risks. Therefore, minimizing unnecessary data collection and applying appropriate access controls are important practices when developing and operating responsible AI solutions.