FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence

dc.contributor.authorSohn, Kihyuk
dc.contributor.authorBerthelot, David
dc.contributor.authorLi, Chun‑Liang
dc.contributor.authorZhang, Zizhao
dc.contributor.authorCarlini, Nicholas
dc.contributor.authorCubuk, Ekin D
dc.contributor.authorKurakin, Alex
dc.contributor.authorZhang, Han
dc.contributor.authorRaffel, Colin
dc.date.accessioned2025-06-02T11:27:40Z
dc.date.available2025-06-02T11:27:40Z
dc.date.issued2020-01-21
dc.descriptionFixMatch는 최소한의 구성으로 준지도 학습을 수행하는 방법으로, 모델의 약한 증강 이미지 예측으로 가짜 레이블 생성 후 강한 증강 이미지 학습에 활용합니다 ©2020 Authors
dc.description.abstractSemi-supervised learning (SSL) provides an effective means of leveraging unlabeled data to improve a model's performance. In this paper, we demonstrate the power of a simple combination of two common SSL methods: consistency regularization and pseudo-labeling. Our algorithm, FixMatch, first generates pseudo-labels using the model's predictions on weakly-augmented unlabeled images. For a given image, the pseudo-label is only retained if the model produces a high-confidence prediction. The model is then trained to predict the pseudo-label when fed a strongly-augmented version of the same image. Despite its simplicity, we show that FixMatch achieves state-of-the-art performance across a variety of standard semi-supervised learning benchmarks, including 94.93% accuracy on CIFAR-10 with 250 labels and 88.61% accuracy with 40 -- just 4 labels per class. Since FixMatch bears many similarities to existing SSL methods that achieve worse performance, we carry out an extensive ablation study to tease apart the experimental factors that are most important to FixMatch's success. We make our code available at this https URL.
dc.description.sponsorshipGoogle Research
dc.identifier.urihttps://arxiv.org/abs/2001.07685
dc.identifier.urihttp://data.inu.ac.kr/handle/123456789/1946
dc.language.isoen_US
dc.publisherarXiv
dc.subjectFixMatch
dc.subjectSemi-Supervised Learning
dc.subjectComputer Vision
dc.subjectSSL
dc.subjectConsistency Regularization
dc.subjectImage analysis
dc.subjectComputer science
dc.titleFixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
dc.title.alternativeFixMatch
dc.typeArticle

Files

Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
2001.07685v2.pdf
Size:
752.12 KB
Format:
Adobe Portable Document Format
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed to upon submission
Description:

Collections