Essay

How can a source ID feature help with mixed-source data, and what is its main drawback?

Question: Suppose a dataset combines product prices collected from several store chains, and the same visible features can correspond to different prices depending on which chain supplied the record. Explain how adding a feature that identifies the source can help, and describe the main practical limitation of this idea.

Sample answer: Adding a source-identifying feature to each example changes the input so that the model can tell where the record came from. With that extra information, the same visible measurements no longer have to map to a single output value, because the source becomes part of the input description. This reduces ambiguity when identical measurements lead to different labels in different sources. A major limitation is that this approach is not commonly used in real applications, so although it is conceptually useful, it is not a standard fix in practice.

Key points:

  • Include a source identifier as part of the input features.
  • The source identifier makes the target less ambiguous for identical observable inputs.
  • The method helps when the same measurements have different labels across sources.
  • A practical drawback is that it is not commonly seen in deployed systems.

Rubric: The response must explain that adding a source identifier to the input can remove ambiguity when the same observable features come from different sources. It must also identify that a practical limitation is that this technique is not widely used in real-world practice.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI

Related