Network Representation Learning

background

Recording studying during KTH. First blog about network representation learning, a project belongs to machine learning, advanced course.

LINE

Reproduce paper “LINE: Large-scale Information Network Embedding”.

Alias Table Method

It’s a method of effiently drawing samples from discrete distribution.
reference:
https://www.keithschwarz.com/darts-dice-coins/
https://blog.csdn.net/haolexiao/article/details/65157026

Negative Sampling

word2vec

Original paper:
Efficient estimation of word representations in vector space.
reference:
word2vec Explained: Deriving Mikolov et al.’s
Negative-Sampling Word-Embedding Method

Skip-Gram Model

Original papaer:Distributed Representations of Words and Phrases
and their Compositionality.
The idea behind the word2vec models is that the words that appear in the same context (near each other) should have similar word vectors. Therefore, we should consider some notion of similarity in our objective when training the model. This is done using the dot product since when vectors are similar, their dot product is larger.
reference:
https://www.baeldung.com/cs/nlps-word2vec-negative-sampling

graphSage