Microsoft AI-900 Online Practice Questions and Exam Preparation

Microsoft AI-900 Online Questions & Answers

• Question 61:

  You need to train a machine learning model by using Azure Machine Learning.

  What is the correct way to prepare data for model evaluation?

  A. Use all data for training only
  B. Split the dataset into training and validation subsets
  C. Use only labeled data for validation
  D. Split the dataset into columns for training and testing
  B. Split the dataset into training and validation subsets

  ---

  Explanation

  Splitting data into training and validation (or test) datasets ensures that the model is evaluated on unseen data. This helps measure how well the model generalizes to new data.

• Question 62:

  You are building a Language Understanding model for an e-commerce business.

  You need to ensure that the model detects when utterances are outside the intended scope of the model.

  What should you do?

  A. Test the model by using new utterances
  B. Add utterances to the None intent
  C. Create a prebuilt task entity
  D. Create a new model
  B. Add utterances to the None intent

  ---

  Explanation

  The None intent is filled with utterances that are outside of your domain.

  References:

  https://docs.microsoft.com/en-us/azure/cognitive-services/LUIS/luis-concept-intent

• Question 63:

  DRAG DROP

  Match the types of computer vision workloads to the appropriate scenarios.

  To answer, drag the appropriate workload type from the column on the left to its scenario on the right. Each workload type may be used once, more than once, or not at all.

  NOTE: Each correct selection is worth one point.

  Select and Place:

  

  

  ---

  Box 1: Image classification

  Image classification is used to determine the main subject of an image. You can use the Azure AI Custom Vision services to train a model that classifies images based on your own categorizations.

  Box 2: Optical Character recognition (OCR)

  Box 3: Object detection

  Object detection is similar to tagging, but the API returns the bounding box coordinates (in pixels) for each object found in the image. For example, if an image contains a dog, cat and person, the Detect operation will list those objects with their coordinates in the image. You can use this functionality to process the relationships between the objects in an image. It also lets you determine whether there are multiple instances of the same object in an image.

  References:

  https://learn.microsoft.com/en-us/training/modules/classify-imageshttps://learn.microsoft.com/en-us/azure/ai-services/computer-vision/concept-object-detection

• Question 64:

  You have the Predicted vs. True chart shown in the following exhibit.

  

  Which type of model is the chart used to evaluate?

  A. classification
  B. regression
  C. clustering
  B. regression

  ---

  Explanation

  What is a Predicted vs. True chart?

  Predicted vs. True shows the relationship between a predicted value and its correlating true value for a regression problem. This graph can be used to measure performance of a model as the closer to the y=x line the predicted values are, the better the accuracy of a predictive model.

  References:

  https://docs.microsoft.com/en-us/azure/machine-learning/how-to-understand-automated-m

• Question 65:

  HOTSPOT

  Select the answer that correctly completes the sentence.

  For _____________ you use a portion of a dataset to prepare a machine learning model and retain the balance of the dataset to verify the results.

  

  

  ---

  During model training, a portion of the dataset (commonly 70-80%) is used to teach the machine learning algorithm to identify patterns and relationships between input features and the output label. The remaining data (usually 20-30%) is held back to evaluate the model's performance and verify its accuracy on unseen data. This ensures the model is not overfitted (too tightly fitted to training data) and can generalize well to new inputs. Key steps highlighted in Microsoft Learn materials:

  Model Training: Use the training data to fit the model - the algorithm learns relationships between input features and labels.

  Model Evaluation: Use the test or validation data to assess the accuracy, precision, recall, or other metrics of the trained model.

  Model Deployment: Once validated, the model is deployed to make real-world predictions. Other options explained: Feature engineering: Involves preparing and transforming input data, not splitting datasets for training and testing. Time constraints: Not a machine learning process step. Feature stripping: Not a recognized ML concept. MLflow models: Refers to an open-source tool for tracking and managing models, not dataset splitting or training. Thus, when you use a portion of the dataset to prepare and train a machine learning model, and retain the rest to verify results, the process is known as model training.

• Question 66:

  A manufacturing company wants to automatically detect labels on products moving along a conveyor belt.

  Which technique should be used?

  A. Image classification
  B. Object detection
  C. Image tagging
  D. Speech recognition
  B. Object detection

  ---

  Explanation

  Object detection identifies specific objects in an image and determines their locations using bounding boxes. This makes it suitable for detecting product labels on moving items. Image classification only assigns a category to the entire image without locating objects.

• Question 67:

  HOTSPOT

  To complete the sentence, select the appropriate option in the answer area.

  

  

  ---

  Accelerate your business processes by automating information extraction. Form Recognizer applies advanced machine learning to accurately extract text, key/value pairs, and tables from documents. With just a few samples, Form Recognizer tailors its understanding to your documents, both on-premises and in the cloud. Turn forms into usable data at a fraction of the time and cost, so you can focus more time acting on the information rather than compiling it.

  Custom Vision

  Form Recognizer

  Ink Recognizer

  Text Analytics References:

  https://azure.microsoft.com/en-us/services/cognitive-services/form-recognizer/

• Question 68:

  You need to identify harmful content in a generative Al solution that uses Azure OpenAI Service.

  What should you use?

  A. Face
  B. Video Analysis
  C. Language
  D. Content Safety
  C. Language

• Question 69:

  You send an image to a Computer Vision API and receive back the annotated image shown in the exhibit.

  

  Which type of computer vision was used?

  A. object detection
  B. semantic segmentation
  C. optical character recognition (OCR)
  D. image classification
  A. object detection

  ---

  Explanation

  Object detection is similar to tagging, but the API returns the bounding box coordinates (in pixels) for each object found. For example, if an image contains a dog, cat and person, the Detect operation will list those objects together with their coordinates in the image. You can use this functionality to process the relationships between the objects in an image. It also lets you determine whether there are multiple instances of the same tag in an image.

  The Detect API applies tags based on the objects or living things identified in the image. There is currently no formal relationship between the tagging taxonomy and the object detection taxonomy. At a conceptual level, the Detect API only finds objects and living things, while the Tag API can also include contextual terms like "indoor", which can't be localized with bounding boxes.

  References:

  https://docs.microsoft.com/en-us/azure/cognitive-services/computer-vision/concept-object-detection

• Question 70:

  You are developing a Language Understanding (LUIS) model.

  What is an utterance?

  A. a category of user intent
  B. a sample user input
  C. a trained model
  D. a dataset
  B. a sample user input

  ---

  Explanation

  An utterance is a phrase or sentence that a user might input, used to train the model to recognize intents.