AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  A type of supervised learning that uses labeled training data

  Why it's wrong here

  This option is incorrect because reinforcement learning is not categorised as a supervised learning method. Supervised learning relies on labelled training data where each example has a known ground-truth outcome, whereas reinforcement learning uses evaluative feedback in the form of rewards and penalties rather than explicit labels. The agent never sees ‘correct’ actions directly; it must discover them through interaction and delayed consequences, which is fundamentally different from mapping inputs to pre-labelled outputs.

• ✓

  Training an agent through rewards and penalties in an interactive environment

  Why this is correct

  This is the correct definition of reinforcement learning. An agent operates in an environment, takes actions, and receives a scalar reward signal—positive for desirable behaviour and negative for undesirable—to learn a policy that maximises cumulative reward over time. This trial-and-error process is formalised as a Markov decision process, and the agent must balance exploration of unknown actions with exploitation of known rewarding ones. Through repeated interactions, the agent learns to associate environmental states with actions that yield the highest long-term return.

• ✗

  A clustering technique that groups similar data automatically

  Why it's wrong here

  This description confuses reinforcement learning with unsupervised clustering. Clustering is an unsupervised technique that partitions data points into groups based on similarity, with no reward signals or sequential decisions involved. Reinforcement learning, by contrast, is driven by an evaluative reward signal from an interactive environment, where the goal is to learn a policy for choosing actions over time—not to discover hidden structures in static data. Thus, while both are ‘learning from experience’ in a loose sense, their objectives and mechanisms are entirely distinct.

• ✗

  Using previously trained models on new tasks

  Why it's wrong here

  This describes transfer learning, which reuses weights or features from a model pre-trained on one task to accelerate or improve learning on a different but related task. Reinforcement learning is not defined by reusing existing models; an agent typically learns from scratch by interacting with its environment and adjusting its policy based on rewards and penalties. Even when prior knowledge is incorporated in RL, the core learning process remains environmental interaction and reward maximisation, not the straightforward application of a pre-trained network to a new task.

Go deeper

Related to this question