Fig.1

Concept

Skip-gram

Skip-gram is a training objective for learning word embeddings by predicting the surrounding context words from a single center word. It flips the more intuitive setup: instead of using neighbors to guess a target (that is CBOW, its sibling in the word2vec paper), Skip-gram…

The rest of “Skip-gram” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.

Log in to unlock

← Back to the library