Embedding

/ɪmˈbɛdɪŋ/em·bed·dingnoun
Representation Learning CoreFoundational

Definition

1.[in machine learning] a low-dimensional vector representation of high-dimensional data, such as words or images, that captures semantic or structural relationships.

We trained a word2vec model to generate a dense vector embedding for each word in the corpus.

Formal statement

v = W x + b

W is the weight matrix, x is the input vector, and b is the bias.

Etymology

Coined by Tomas Mikolov in 2013, derived from the general concept of mapping complex data into a continuous vector space.

Synonyms

  • vector representationsense 1 · Absolute

References

  • Mikolov et al. (2013). Efficient Estimation of Word Representations in Vector Space.NIPS Workshop.