Skip to content

FRM Part I · FRM Exam Part I · Machine-Learning Methods

A trading desk builds an algorithm that repeatedly chooses how much of a position to execute each minute. After each action it observes the resulting transaction cost and adjusts its strategy to lower cumulative cost over time, with no labeled correct actions supplied. Which description is most accurate?

This is reinforcement learning. The algorithm takes sequential actions, receives cost feedback as a penalty or reward, and adapts to minimize cumulative cost without labeled correct actions. Supervised regression needs fixed input-target pairs, and clustering only groups unlabeled data without optimizing rewards.

  1. AReinforcement learning, because the agent learns from rewards or penalties tied to its sequential actionsCorrect
  2. BSupervised regression, because transaction cost is a numeric target
  3. CUnsupervised clustering, because no correct actions are labeled
  4. DSupervised classification, because each action is one of a discrete set

Explanation

The agent interacts with an environment, takes sequential actions and receives feedback in the form of costs (negative rewards), aiming to optimize cumulative reward. That is reinforcement learning. Cost is feedback to actions, not a fixed label paired with inputs, so supervised regression is wrong, and clustering seeks no reward optimization.

Did you get it right without looking?

One question tells you little. A timed set on Machine-Learning Methods shows your real accuracy, how long you take and where you lose marks.

Practise Machine-Learning MethodsFree diagnostic mock

More Machine-Learning Methods questions