Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Euclidean distance is a measure that may be used to compute the similarity between two vectors. Given a query $q$ and documents $d_1, \ldots, d_n$, we may rank the documents $\mathcal{D} = \{d_k\}_{k=1}^n$ in the increasing order of Euclidean distance from $q$.

Prove that if $q$ and the document vectors $\mathcal{D}$ are all normalized to unit vectors, then the rank ordering produced by Euclidean distance is identical to that produced by cosine similarity.

share|improve this question
What is a cosine similarity? Computing $\dfrac{\vec{x}\cdot \vec{y}}{\|\vec{x}\| \|\vec{y}\|}$? –  gt6989b Mar 20 '13 at 17:10
@gt6989b Yes en.wikipedia.org/wiki/Cosine_similarity#Definition –  benji_r Mar 20 '13 at 17:11
add comment

1 Answer 1

Doesn't seem true to me. Take $\mathcal{D}=\{(1,0), (0,1)\}$ and $q = (a,b)$ on the unit circle. Then, the cosine similarity would result in $q \cdot d_1 = a$ and $q \cdot d_2 = b$. Euclidean distance (we consider the square of the actual distance, it doesn't alter comparisons), however, would have $(a-1)^2+b^2$ and $a^2 + (b-1)^2$.

Note that $$ \begin{split} (a-1)^2+b^2 &< a^2 + (b-1)^2 \\ a^2 -2a + 1+b^2 &< a^2 + b^2 -2b +1 \\ -2a &< -2b \\ b &< a \end{split} $$

So the distance ranking orders $d_1 < d_2$ iff the cosine ranking orders $d_2 < d_1$.

It seems, this is the counterexample, since they produce a different ordering. Or am I not understanding something?

share|improve this answer
add comment

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.