Machine Learning::Embedding

A good representation should be able to separate different instances and cluster similar instances.

Use the context to predict the center word

negative sampling makes the calculations faster

Use the center word to predict the context