Embedding layer Deep Learning
noun phrase
Definition: A neural-network layer that maps discrete indices (e.g., token IDs) to dense continuous vectors (embeddings) [Keras Documentation; TensorFlow Docs].
Example in context: “CNNs can either receive as input pre-trained embeddings (such as those from a pre-trained word2vec model) or train their own embeddings by using what is known as an embedding layer.” [Pereira et al. 2024]
Related terms: token embedding layer, lookup layer, word embedding layer