#one-hot-encoding

Question 1: Single choice

A customer table has a categorical column customer_id whose value differs in every row. You one-hot encode all categorical columns before training a model. What does this column do?

One indicator per row, which lets the model memorize rows but tells it nothing about new customers1
It adds a single numeric column, because a column whose values are all unique is treated as an integer feature2
It becomes the most informative feature, because it identifies every training example exactly3
It is harmless, because a regularized model automatically ignores columns that do not help4

Timed test