AnnoyIndexer() approximate_neighbors with the cosine similarity or Angular distance ?

[SAma TIma]

hello every one ! 

as the title said : when i see in the documentation for annoy ,  Annoy uses Euclidean distance of normalized vectors for its angular distance which for two vectors u,v is equal to 2(1-cos(u,v)) ,

and in the tutorial  they said :the approximate_neighbors and their corresponding cosine similarity values

https://markroxor.github.io/gensim/static/notebooks/annoytutorial.html

is it the same thing and i can't understand or what? 

thank you 

[Ivan Menshikh]

Hello,

AnnoyIndexer used the default value of "metric" mentioned in https://github.com/spotify/annoy. This is possible to use different (if you use pure-annoy without AnnoyIndexer), but only one metric for AnnoyIndexer.