how does n_similarity work?

[Joan Carrillo Baron]

Hello Gensim group I am training a model Word2Vec, but I would like to know how n_similarity works.  I know it works with cosine distance, but I want to deepen.

I normally send a matrix of vectors with 250 dimensions each.

Thank u very much

[Gordon Mohr]

The source code is 5 lines:

https://github.com/RaRe-Technologies/gensim/blob/a435f24fe25e17f473e71af3468660512c2606cb/gensim/models/keyedvectors.py#L1239

The doc-comment is somewhat out-of-date; as the helper method `KeyedVectors.get_mean_vector()` now accepts either keys to lookup known vectors, or raw vectors themselves, the method can do more than the description "Compute cosine similarity between two sets of keys." implies. But, if you're already sending it vectors, you already know that. 

What remains unclear, specifically?

- Gordon