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

 

Question 201. What is the main purpose of a machine-learning model registry?

1. To manage and track registered model versions
2. To translate documents
3. To store network passwords
4. To analyze audio

Correct Answer: 1. To manage and track registered model versions

Explanation:

A model registry provides a centralized way to manage machine-learning models and their versions throughout the development and deployment lifecycle. It can help teams keep track of which models have been trained, evaluated, and made available for deployment. Versioning is particularly important because developers may create multiple versions of a model as data, features, algorithms, or training procedures change. A registry can make it easier to identify a particular model version and understand its associated metadata. It is not primarily designed for translation, audio analysis, or password storage. Model management and version tracking are therefore key purposes of a model registry.

Question 202. Which component identifies the expected output in a supervised learning dataset?

1. Feature
2. Label
3. Embedding
4. Endpoint

Correct Answer: 2. Label

Explanation:

A label represents the expected target or outcome that a supervised machine-learning model is trained to predict. For example, in a dataset used to predict whether a customer will cancel a subscription, the label could indicate whether cancellation occurred. Features provide the input information that the model uses to make its prediction. During training, the algorithm learns relationships between the features and the known labels. Embeddings are numerical representations commonly used in AI applications, while an endpoint is associated with accessing a deployed service or model. Therefore, the expected output in supervised learning is represented by the label.

Question 203. Which approach can help improve the quality of training data for an AI model?

1. Removing all useful examples
2. Ignoring missing values
3. Using representative and relevant data
4. Reducing the dataset without evaluation

Correct Answer: 3. Using representative and relevant data

Explanation:

Using representative and relevant training data can improve an AI model because the model learns patterns from the examples provided during training. If the training dataset accurately reflects the situations, users, and conditions that the model will encounter, the resulting model has a better opportunity to generalize to real-world inputs. Data quality also involves checking for missing values, incorrect records, duplicates, and inappropriate information. Simply removing useful examples or ignoring data-quality problems can reduce model effectiveness. Reducing the dataset may sometimes be appropriate, but it should be based on a clear reason and evaluation. Representative data is therefore an important consideration in model development.

Question 204. What is a major risk of using a dataset that does not represent the intended user population?

1. The model may perform differently across groups
2. The model automatically becomes faster
3. The storage system becomes encrypted
4. The application requires no testing

Correct Answer: 1. The model may perform differently across groups

Explanation:

A dataset that does not adequately represent the intended population can cause an AI model to perform differently for different groups or situations. If certain populations are missing or underrepresented in training data, the model may not learn patterns that apply to those users effectively. This can create fairness and performance concerns when the system is deployed. Representative data is therefore important when developing and evaluating AI systems. Testing across relevant groups can also help identify differences in performance. Data imbalance does not automatically make an application faster, encrypt storage, or eliminate the need for testing. Data representation should be considered throughout the AI lifecycle.

Question 205. Which feature of Azure Machine Learning helps organize reusable data used by machine-learning workflows?

1. Data assets
2. Firewall rules
3. DNS records
4. Virtual network gateways

Correct Answer: 1. Data assets

Explanation:

Azure Machine Learning data assets provide a managed way to register and reference data used by machine-learning workflows. Organizing data as assets can make it easier for teams to reuse datasets, track data versions, and reference consistent data in training or other jobs. This is useful in repeatable machine-learning processes where different experiments or pipelines need access to known data resources. Firewall rules, DNS records, and virtual network gateways address network or infrastructure requirements rather than data management within machine-learning workflows. Data assets therefore help bring structure and traceability to the data used by Azure Machine Learning solutions.

Question 206. What is the purpose of hyperparameters in machine learning?

1. They store user passwords
2. They control aspects of the training process
3. They replace the training dataset
4. They identify network devices

Correct Answer: 2. They control aspects of the training process

Explanation:

Hyperparameters are configuration values selected before or during model training that influence how the learning algorithm operates. Examples can include learning rate, number of trees, maximum tree depth, batch size, or the number of training iterations, depending on the algorithm. Unlike model parameters, which are learned from the training data, hyperparameters are generally configured by the developer or selected through a tuning process. Appropriate hyperparameter choices can influence model performance and training efficiency. They do not store passwords, replace datasets, or identify network devices. Understanding hyperparameters is important because different configurations can produce significantly different model results.

Question 207. Which technique can automatically test different machine-learning algorithms and configurations to identify a suitable model?

1. Automated Machine Learning
2. OCR
3. Speech synthesis
4. DNS resolution

Correct Answer: 1. Automated Machine Learning

Explanation:

Automated Machine Learning, commonly called Automated ML, can automate parts of the model-development process by trying different algorithms, preprocessing approaches, and parameter configurations for a specified machine-learning task. It can evaluate candidate models using selected metrics and help identify promising configurations. This can reduce the amount of manual experimentation required when developing certain supervised-learning solutions. Automated ML does not mean that every aspect of an AI project is automatically completed; data quality, business requirements, responsible AI considerations, and final validation still require appropriate attention. OCR, speech synthesis, and DNS resolution provide unrelated capabilities.

Question 208. What is the purpose of a test dataset in machine learning?

1. To configure model hyperparameters repeatedly
2. To provide an independent final evaluation of the model
3. To replace all training data
4. To create network security rules

Correct Answer: 2. To provide an independent final evaluation of the model

Explanation:

A test dataset is generally reserved for evaluating a trained model after development and tuning decisions have been completed. Because the test data is kept separate from the training and development process, it can provide a more independent indication of how the model may perform on unseen data. For example, a classification model can be evaluated on test records that were not used to fit the model or select its hyperparameters. Using the test dataset repeatedly during development can reduce its independence. Therefore, its primary purpose is to provide a final or unbiased evaluation rather than replace training data or configure network security.

Question 209. Which Azure service can analyze the visual content of images?

1. Azure AI Vision
2. Azure AI Translator
3. Azure DNS
4. Azure Key Vault

Correct Answer: 1. Azure AI Vision

Explanation:

Azure AI Vision provides computer vision capabilities for analyzing visual information. Depending on the capability being used, an application can identify objects, extract text, generate descriptions, analyze image content, or perform other supported vision tasks. For example, an application could analyze uploaded photographs and identify visual features or extract text using optical character recognition. Azure AI Translator is intended for language translation, Azure DNS provides domain name resolution, and Azure Key Vault helps manage secrets and cryptographic keys. Therefore, Azure AI Vision is the appropriate service when the requirement involves analyzing the visual content of images.

Question 210. Which computer vision task identifies objects and their locations within an image?

1. Sentiment analysis
2. Object detection
3. Language detection
4. Text summarization

Correct Answer: 2. Object detection

Explanation:

Object detection identifies objects within an image and determines where those objects are located, commonly by returning bounding boxes around detected items. For example, an application analyzing a street photograph might identify several cars, people, and traffic signs and provide the locations of those objects. This differs from image classification, which assigns categories to an image without necessarily locating each object. Sentiment analysis and language detection are language-related capabilities, while text summarization reduces written content into a shorter representation. Therefore, object detection is the appropriate computer vision task when both the presence and location of objects need to be identified.

Question 211. Which task separates different regions or objects at the pixel level in an image?

1. Image segmentation
2. Speech-to-text
3. Regression
4. Translation

Correct Answer: 1. Image segmentation

Explanation:

Image segmentation divides an image into meaningful regions, often assigning a category or object identity to individual pixels. This provides more detailed information than simply drawing a bounding box around an object. For example, segmentation can be useful when an application needs to identify the precise shape or area occupied by an object in an image. Object detection identifies objects and their approximate locations, while segmentation provides more detailed pixel-level information. Speech-to-text processes spoken audio, regression predicts numerical values, and translation converts content between languages. Therefore, image segmentation is the appropriate technique when precise regions or object boundaries are required.

Question 212. Which Azure AI Language capability can identify the language used in a text input?

1. Key phrase extraction
2. Language detection
3. Object detection
4. OCR

Correct Answer: 2. Language detection

Explanation:

Language detection determines which language or languages are represented in a given text input. This capability is useful for multilingual applications where the system needs to understand the input language before selecting an appropriate processing workflow. For example, a customer-support application may receive messages in several languages and use language detection to route each message to the appropriate translation or language-processing process. Key phrase extraction instead identifies important concepts or phrases in text. Object detection works with images, while OCR extracts visible text from images or documents. Therefore, language detection is the capability specifically designed to identify the language of text.

Question 213. Which natural language processing capability extracts important terms from a document?

1. Key phrase extraction
2. Object detection
3. Image segmentation
4. Speech synthesis

Correct Answer: 1. Key phrase extraction

Explanation:

Key phrase extraction identifies important words or phrases that represent the main concepts within a text document. It can help summarize the subject matter of large amounts of unstructured text without requiring a complete manual review. For example, a customer-feedback document might produce key phrases related to delivery delays, product quality, customer service, or pricing. This capability can be useful for organizing documents, analyzing feedback, and supporting search or analytics workflows. Object detection and image segmentation are computer vision tasks, while speech synthesis converts text into spoken audio. Therefore, key phrase extraction is the appropriate capability when an application needs to identify important terms in text.

Question 214. Which Azure AI capability can convert a written document into a structured set of extracted fields?

1. Azure DNS
2. Azure AI Document Intelligence
3. Azure Firewall
4. Azure Virtual Network

Correct Answer: 2. Azure AI Document Intelligence

Explanation:

Azure AI Document Intelligence can analyze supported documents and extract information into structured outputs. This capability is useful for documents such as invoices, receipts, forms, and other business records where the application needs to identify fields and their corresponding values. Rather than requiring a developer to manually interpret every document layout, document-analysis capabilities can recognize relevant information and return it in a form that applications can process. Azure DNS, Azure Firewall, and Azure Virtual Network are infrastructure services with different purposes. Therefore, Document Intelligence is appropriate when the requirement is to transform information contained in documents into structured data.

Question 215. Which Azure service provides machine translation capabilities for multilingual applications?

1. Azure Monitor
2. Azure AI Translator
3. Azure Blob Storage
4. Azure Machine Learning compute

Correct Answer: 2. Azure AI Translator

Explanation:

Azure AI Translator provides machine translation capabilities that can be incorporated into applications requiring multilingual communication. A solution can submit text in a supported source language and request translation into a supported target language. This can be useful for websites, customer-support applications, document workflows, and other systems where users may communicate in different languages. Azure Monitor focuses on monitoring resources and applications, Azure Blob Storage provides object storage, and Azure Machine Learning compute supplies processing resources for machine-learning workloads. Therefore, Azure AI Translator is the service specifically intended to provide automated translation between supported languages.

Question 216. Which generative AI concept refers to the instructions or input provided to a language model?

1. Prompt
2. Label
3. Feature
4. Endpoint

Correct Answer: 1. Prompt

Explanation:

A prompt is the input or set of instructions provided to a generative AI model to guide the response it produces. Prompts can contain questions, task descriptions, context, examples, constraints, and other instructions. The quality and structure of a prompt can influence how effectively the model understands the intended task. For example, a developer might provide a prompt asking a model to summarize a document using a specific format and audience level. Labels and features are primarily machine-learning dataset concepts, while an endpoint is an interface through which an application can access a deployed model or service. Therefore, the input instructions are referred to as a prompt.

Question 217. What is zero-shot prompting?

1. Training a model with zero records
2. Providing a model with no examples while asking it to perform a task
3. Removing all model parameters
4. Running only batch inference

Correct Answer: 2. Providing a model with no examples while asking it to perform a task

Explanation:

Zero-shot prompting asks a generative AI model to perform a task without providing explicit examples of the desired input-output pattern. The prompt describes the task and may include relevant instructions or context, but it does not include demonstration examples. For instance, a developer could ask a language model to classify a customer comment as positive or negative while simply defining the categories rather than showing example classifications. This differs from few-shot prompting, where several examples are included. Zero-shot prompting does not mean that the model has no training data; it means that no task-specific examples are supplied in the prompt.

Question 218. Which practice can help make a generative AI prompt more precise?

1. Removing all task requirements
2. Providing clear instructions and relevant context
3. Using unrelated information
4. Avoiding any description of the desired output

Correct Answer: 2. Providing clear instructions and relevant context

Explanation:

Clear instructions and relevant context can help a generative AI model understand the intended task and produce a more useful response. A well-designed prompt can specify the goal, audience, required format, constraints, available information, and desired characteristics of the output. For example, instead of asking a model to “write something,” a developer could specify that it should summarize a document in five concise bullet points for a business audience. Removing task requirements or adding unrelated information can make the request less clear. Describing the expected output format can also improve consistency. Prompt engineering therefore involves intentionally structuring inputs to guide model behavior.

Question 219. Which responsible AI principle focuses on ensuring an AI system works reliably and handles failures appropriately?

1. Reliability and safety
2. Translation
3. Data compression
4. Clustering

Correct Answer: 1. Reliability and safety

Explanation:

Reliability and safety focus on ensuring that an AI system performs consistently and can operate safely within its intended environment. AI applications should be designed with their expected conditions, possible failures, and potential risks in mind. Testing, monitoring, validation, safeguards, and appropriate error handling can all contribute to more reliable systems. For example, an application may need to recognize when it cannot confidently complete a task and provide an appropriate fallback rather than producing an unsupported result. Translation and clustering are AI capabilities or techniques, while data compression is a storage-related process. Therefore, reliability and safety is the responsible AI principle associated with dependable and safe operation.

Question 220. Why is it important to define the intended use of an AI system before deployment?

1. It removes the need for evaluation
2. It guarantees perfect accuracy
3. It helps determine appropriate requirements, risks, and safeguards
4. It prevents all future data changes

Correct Answer: 3. It helps determine appropriate requirements, risks, and safeguards

Explanation:

Defining the intended use of an AI system helps developers and organizations understand what the system is expected to do, who will use it, what data it will process, and what risks may need to be addressed. These considerations influence model selection, evaluation criteria, security controls, human oversight, and responsible AI safeguards. For example, an AI system used for a low-risk content recommendation scenario may require different controls from one used to support an important business decision. Defining intended use cannot guarantee perfect accuracy or prevent future changes in data. Instead, it establishes a clear context for designing, testing, deploying, and monitoring the AI solution.