| 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
- BYOL
- CGAN
- simclrv2
- remixmatch paper
- UnderstandingDeepLearning
- mme paper
- 최린컴퓨터구조
- CoMatch
- shrinkmatch
- mocov3
- adamatch paper
- dcgan
- tent paper
- SSL
- semi supervised learnin 가정
- 컴퓨터구조
- shrinkmatch paper
- cifar100-c
- Pseudo Label
- WGAN
- Entropy Minimization
- dann paper
- CycleGAN
- ConMatch
- Pix2Pix
- conjugate pseudo label paper
- 딥러닝손실함수
- GAN
- 백준 알고리즘
- Meta Pseudo Labels
Archives
- Today
- Total
목록CCSSL (1)
Hello Data
논문의 풀 제목은 Class-Aware Contrastive Semi Supervised Learning이다. https://arxiv.org/pdf/2203.02261.pdf Introduction 모델을 훈련시킬 때 중요한 요소가 있는데 바로 Raw data에 대한 접근이다. Real world에서는 Raw data가 많을 것이며 이 중 일부는 훈련에서 경험하지 못한 클래스들의 데이터들도 있을 것이다. 이것을 out of distribution data라고 하며 훈련과정에서 경험한 데이터를 in distribution data라고 한다. 이를 잘 보여주는 사진이다. SSL을 통해 raw data 에 대한 대비 없이 훈련하게 된다면 confirmation bias, 즉 확증편향이 일어나게 되는데 이는 ..
Self,Semi-supervised learning
2023. 8. 2. 18:44