Adaptive Skip-gram (AdaGram) model is a nonparametric extension of famous Skip-gram model implemented in word2vec software which is able to learn multiple representations per word capturing different word meanings. This is C++ ported version of AdaGram in Julia (https://github.com/sbos/AdaGram.jl).
You can see more informations about AdaGram at this repository(https://github.com/sbos/AdaGram.jl).
- Eigen (C++ Library) : http://eigen.tuxfamily.org
By using Eigen, This code improves performance up to ~x2 compared to original Julia version, ~x3 when compiling with SSE2, and ~x5 when compiling with AVX.