Prepare and test your skills
Prepare and test your skills
Free sample question, answered and explained below. The Generative AI Leader practice bank has 2,191 more like it. Try 10 free: instant score, no account.
A retail company wants to train a machine learning model to automatically route incoming customer support emails to the correct department, such as Billing, Technical Support, or Shipping.
The project team compiles a dataset containing thousands of past customer emails, where each email is explicitly paired with the verified department tag that resolved the issue.
Which statement correctly describes this dataset and the learning method used?
Reinforcement learning (RL) is a machine learning paradigm where an autonomous agent learns to make a sequence of decisions by interacting with a dynamic environment. Rather than learning from static labeled datasets, the agent observes the current state, executes an action, and receives feedback in the form of numerical rewards or penalties, iteratively optimizing a policy to maximize cumulative rewards over time.
Unlabeled data consists of raw inputs (such as text documents, images, or sensor logs) that have no associated target tags, category names, or ground truth annotations. Unsupervised learning algorithms ingest this unlabeled data to discover hidden patterns, underlying structures, anomaly distributions, or natural groupings (clustering) on their own without human guidance on the correct answer.
Validation data is a specific subset of a curated dataset separated from training data to evaluate model performance, tune hyperparameters, and detect overfitting during development. Heuristic rules (rule-based programming) rely on manually written if-then logical statements constructed by human programmers rather than statistical patterns learned by machine learning models.
Labeled data consists of raw input samples (such as text, images, or audio) paired with an explicit target tag, annotation, or ground truth. In machine learning, ground truth represents the verified, correct answer or outcome for each specific example. Supervised learning is the training methodology where an algorithm is provided with these labeled input-output pairs so it can learn the mathematical relationships and patterns connecting features to outcomes.

Supervised learning on labeled data is the most reliable approach when the desired output classes are well-defined in advance. Because the company already has verified historical examples of emails mapped to specific departments, providing these ground truth pairs allows the model to optimize its parameters directly for accurate category prediction.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated