問題文
A report needs the ten nearest rows by euclidean distance and does not display the distance values themselves. The team wants the cheapest calculation that keeps the same ten rows in the same order. Which choice meets that requirement?
選択肢
- Use the euclidean squared metric because dropping the square root keeps the relative order of the rows unchanged.
- Use the manhattan metric because summing the differences per dimension keeps the same ten rows in the euclidean order.
- Use the cosine metric because the angle between two vectors carries the same information as their euclidean distance.
- Use the dot metric because the negated inner product ranks the rows in the euclidean order at a lower cost.