https://en.wikipedia.org/wiki/Word2vec
PS. 看文件時要先確認一下是Skip-gram / CBOW / LM 的哪一種, 不然會混淆
input one-hot
output simularity(機率)
只有一層hidden layer
用skip-gram或CBOW概念, 去自訓練出hidden layer前後Weight Matrix
重點是前面的Weight Matrix, train完之後, 每個row即word embedding
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment