Question: In a supervised learning setting, imagine you have a main labeled dataset and an additional labeled dataset collected from a different process. Explain when the two sources are considered consistent. In your explanation, describe the meaning of a single mapping f(x) from input x to label y, and name the main practical drawback of training with the extra source.

Sample answer: The two sources are consistent if one prediction rule can be used for both of them: the same function f(x) should map an input x to the correct label y no matter which source produced the example. The main downside of using the extra source is higher training cost, since the model must process more examples and therefore use more computation.

Key points:

• Consistency means one input-to-label rule works across both sources.
• f(x) should predict y without needing to know where the example came from.
• The main practical drawback is added computation from training on more data.

Rubric: Grading Rubric:

• Describes consistency as a shared mapping from inputs to labels across both sources.
• States that the function f(x) does not depend on the source identity.
• Identifies increased computational cost as the primary practical disadvantage.