the doc of gensim.matutils.kullback_leibler is wrong?

[周唐]

gensim.matutils.kullback_leibler(vec1, vec2, num_features=None)

A distance metric between two probability distributions. Returns a distance value in range <0,1> where values closer to 0 mean less distance (and a higher similarity) Uses the scipy.stats.entropy method to identify kullback_leibler convergence value. If the distribution draws from a certain number of docs, that value must be passed.

but I cal some num bigger than 1, and in wiki:

 

• The Kullback–Leibler divergence is always non-negative,

{\displaystyle D_{\mathrm {KL} }(P\|Q)\geq 0,\,}

a result known as Gibbs' inequality, with D_KL(P‖Q) zero if and only if P = Q almost everywhere. The entropy H(P) thus sets a minimum value for the cross-entropy H(P,Q), the expected number of bits required when using a code based on Q rather than P; and the Kullback–Leibler divergence therefore represents the expected number of extra bits that must be transmitted to identify a value x drawn from X, if a code is used corresponding to the probability distribution Q, rather than the "true" distribution P.

 Wrong?

[Lev Konstantinovskiy]

Hi,

Thanks for the catch. It's indeed an error. Would appreciate a small PR on github.