Machine Learning: Embedding

发表于 阅读次数： 本文字数： 417 阅读时长 ≈ 1 分钟

Embedding

Embedding is a function which input discrete variables and output continuous vectors. It can reduce the dimensionality of the input variables. Such that inputs that behave similarly have similar embedding outputs.

Neural network embeddings have 3 primary purposes:

  1. Fidding nearest neighbors in the embedding space. 这个有点像聚类
  2. As input to a machine learning model.
  3. For visualization between concepts and categories. 可以将对应的输入和embedding情况结合已知的分类条件进行可视化的观察。

How to Get Embedding