Question: Explain why creating a machine learning system usually happens in repeated cycles instead of one straight pass. What are the usual stages in that cycle?

Sample answer: Machine learning work is cyclical because a first solution is rarely good enough on the first try. A team typically begins by choosing one idea to test. They then turn that idea into a working implementation. Next, they run an evaluation to see what happened. The results often show flaws, unexpected behavior, or room for improvement, so the team uses those results to choose a better next idea. The process continues in this way until the system reaches an acceptable level of performance.

Key points:

• The common sequence is plan, implement, and evaluate.
• An early attempt usually falls short of the target.
• Evaluation results guide the next round of changes.
• The cycle is repeated many times before the solution is satisfactory.

Rubric: A full-credit response must name the plan -> implement -> evaluate cycle, state that early attempts usually are not sufficient, and explain that results from evaluation are used to shape the next idea in an ongoing loop.