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 | 29 | 30 |
Tags
- Skip connection
- EDA 추천 파이썬
- EDA in python
- pandas
- 비선형함수
- molecular representation
- 일부 레이어 고정
- sktime tutorial
- pretraining
- pytorch dataset split
- layer 일부 freeze
- weight 일부 고정
- pandas 조건
- 선형함수 딥러닝
- 판다스 조건
- 비선형함수 딥러닝
- pytorch 데이터셋 나누기
- Graph Theory
- sktime
- 시계열 라이브러리
- sktime 예제
- python 경우의 수
- sktime 튜토리얼
- Does GNN Pretraining Help Molecular Representation?
- 경우의 수 파이썬
- pandas 특정 조건 열 제거
- Prompt Tuning for Graph Neural Networks
- pandas row 제거
- pandas 행 제거
- 모델 freeze
Archives
- Today
- Total
MoonNote
MolCLR: Molecular Contrastive Learning of Representations via Graph Neural Networks 본문
Study/Paper Review
MolCLR: Molecular Contrastive Learning of Representations via Graph Neural Networks
Kisung Moon 2021. 12. 23. 17:423줄 요약
- molecular 분야는 labeled data가 부족
- contrastive를 위한 augmentation 방법으로 atom masking, bond deletion, subgraph removal 제안
- Encoder로 5-layer GIN 사용
Contrastive Learning


- SimCLR에서 제안한 NT-Xent (normalized temperature-scaled cross entropy) loss를 loss function 으로 사용
- cosine similarity 활용

Results

Comment
- 기존 Graph Contrastive Learning에서 Graph를 Molecule로 치환한 Idea 말고는 큰 novelty가 없는 것 같다.
- augmenation 방법도 기존에 제안했던 방법과 거의 동일하고 node를 atom으로, edge를 bond로 용어를 바꾼 것 말고는 새로운 방법은 아니라고 생각된다.
'Study > Paper Review' 카테고리의 다른 글
Comments