4
$\begingroup$

For a research report, I want to show that the standard expression of the Hellinger Distance between two discrete distributions,

$$H(p,q)={\sqrt{\frac{1}{2} \sum_{x \in X} \left[\sqrt{p(x)}-\sqrt{q(x)} \right]^{2}}}$$

is equivalent to the alternative expression

$$H(p,q)={\sqrt{1-\sum_{x \in X} \sqrt{p(x)q(x)}}}$$

I have tried the expand the square but can't seem to equate the two expressions. Can anyone help?

$\endgroup$
5
  • 1
    $\begingroup$ @User1865345 - I take that to equal 1 $\endgroup$
    – anna6931
    Oct 3, 2022 at 8:59
  • $\begingroup$ @User1865345 - Bingo! I see it now - I got it to equate! Thanks so much. $\endgroup$
    – anna6931
    Oct 3, 2022 at 9:19
  • $\begingroup$ I have turned that into answer. $\endgroup$ Oct 3, 2022 at 9:24
  • $\begingroup$ A remark. There exist different versions of the formula of Hellinger distance, they differ by a constant factor. The version I've seen more often uses $2$ instead of $\frac{1}{2}$ under the root. Note also that if you remove this coefficient (set it to $1$), that will be distance sometimes called Matusita distance. Ref: Michel Deza, Elena Deza. Encyclopedia of Distances. Also: Sung-Hyuk Cha. Comprehensive Survey on Distance/Similarity Measures between Probability Density Functions // INTERNATIONAL JOURNAL OF MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES $\endgroup$
    – ttnphns
    Oct 3, 2022 at 15:30
  • 2
    $\begingroup$ @ttnphns - I have a few references on the Hellinger Distance (43 at last count) and I have only come across two references where the constant of $2$ is used (which are the two you list) hence this seems to be a non-standard version. The version with the $\frac{1}{2}$ factor, as given by the OP, seems to be more common, most likely because the metric is bounded by [0,1]. (Note: Re-posted comment to correct a typographical error). $\endgroup$
    – Mari153
    Oct 4, 2022 at 8:09

1 Answer 1

7
$\begingroup$

Hint: what would $\sum_{x\in \mathcal X} p(x) $ be equal to?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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