Notice
Recent Posts
Recent Comments
Link
일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | ||||||
2 | 3 | 4 | 5 | 6 | 7 | 8 |
9 | 10 | 11 | 12 | 13 | 14 | 15 |
16 | 17 | 18 | 19 | 20 | 21 | 22 |
23 | 24 | 25 | 26 | 27 | 28 |
Tags
- text
- cs224n
- computer
- nlp
- natural_language_processing
- Ai
- slideshare
- Statistics
- cs231n
- Vim
- machinelearning
- language_model
- computer_setting
- Stanford
- deeplearning
- gensim
- tab
- github
- pip
- paper_review
- git
- terminal
- seq2seq
- linux
- install
- code
- json
- review
- Standford
- error
Archives
- Today
- Total
목록Attention (1)
NLP/AI/Statistics
![](http://i1.daumcdn.net/thumb/C150x150/?fname=https://blog.kakaocdn.net/dn/dRb7dY/btqZdlj7l8F/vWaEbovupDcQdk9KKks7Z1/img.png)
앞서 Seq2Seq에 대하여 설명했다. Seq2Seq는 encoder, decoder 개념을 사용하여 입력 시퀀스를 압축하고 출력 시퀀스를 생성하는 모델이다. 이는 기계번역 task에서 주로 사용되지만, encoder에서 일정 크기로 모든 시퀀스 정보를 압축하여 표현하려고 하기 때문에 정보손실이 발생하는 문제가 있다. 이러한 문제를 보완하고자 Attention 모델이 제안되었다. Attention 모델은 decoder가 단순히 encoder의 압축된 정보만을 받아 예측 시퀀스를 출력하는 것이 아니라, decoder가 출력되는 시점마다 encoder에서의 전체 입력 문장을 한번더 검토하도록 한다. 이 때 decoder는 encoder의 모든 입력 시퀀스를 동일한 가중치로 받아들이지 않고, 중요한 단어에 대..