Matches in SemOpenAlex for { <https://semopenalex.org/work/W4281681567> ?p ?o ?g. }
Showing items 1 to 80 of
80
with 100 items per page.
- W4281681567 abstract "Deep neural networks achieve remarkable performances on a wide range of tasks with the aid of large-scale labeled datasets. Yet these datasets are time-consuming and labor-exhaustive to obtain on realistic tasks. To mitigate the requirement for labeled data, self-training is widely used in semi-supervised learning by iteratively assigning pseudo labels to unlabeled samples. Despite its popularity, self-training is well-believed to be unreliable and often leads to training instability. Our experimental studies further reveal that the bias in semi-supervised learning arises from both the problem itself and the inappropriate training with potentially incorrect pseudo labels, which accumulates the error in the iterative self-training process. To reduce the above bias, we propose Debiased Self-Training (DST). First, the generation and utilization of pseudo labels are decoupled by two parameter-independent classifier heads to avoid direct error accumulation. Second, we estimate the worst case of self-training bias, where the pseudo labeling function is accurate on labeled samples, yet makes as many mistakes as possible on unlabeled samples. We then adversarially optimize the representations to improve the quality of pseudo labels by avoiding the worst case. Extensive experiments justify that DST achieves an average improvement of 6.3% against state-of-the-art methods on standard semi-supervised learning benchmark datasets and 18.9%$ against FixMatch on 13 diverse tasks. Furthermore, DST can be seamlessly adapted to other self-training methods and help stabilize their training and balance performance across classes in both cases of training from scratch and finetuning from pre-trained models." @default.
- W4281681567 created "2022-06-13" @default.
- W4281681567 creator A5018357165 @default.
- W4281681567 creator A5019241553 @default.
- W4281681567 creator A5034579434 @default.
- W4281681567 creator A5058280062 @default.
- W4281681567 creator A5059586503 @default.
- W4281681567 creator A5077395936 @default.
- W4281681567 date "2022-02-14" @default.
- W4281681567 modified "2023-09-27" @default.
- W4281681567 title "Debiased Self-Training for Semi-Supervised Learning" @default.
- W4281681567 doi "https://doi.org/10.48550/arxiv.2202.07136" @default.
- W4281681567 hasPublicationYear "2022" @default.
- W4281681567 type Work @default.
- W4281681567 citedByCount "0" @default.
- W4281681567 crossrefType "posted-content" @default.
- W4281681567 hasAuthorship W4281681567A5018357165 @default.
- W4281681567 hasAuthorship W4281681567A5019241553 @default.
- W4281681567 hasAuthorship W4281681567A5034579434 @default.
- W4281681567 hasAuthorship W4281681567A5058280062 @default.
- W4281681567 hasAuthorship W4281681567A5059586503 @default.
- W4281681567 hasAuthorship W4281681567A5077395936 @default.
- W4281681567 hasBestOaLocation W42816815671 @default.
- W4281681567 hasConcept C112972136 @default.
- W4281681567 hasConcept C119857082 @default.
- W4281681567 hasConcept C121332964 @default.
- W4281681567 hasConcept C13280743 @default.
- W4281681567 hasConcept C136389625 @default.
- W4281681567 hasConcept C153180895 @default.
- W4281681567 hasConcept C153294291 @default.
- W4281681567 hasConcept C154945302 @default.
- W4281681567 hasConcept C15744967 @default.
- W4281681567 hasConcept C185798385 @default.
- W4281681567 hasConcept C205649164 @default.
- W4281681567 hasConcept C2776145971 @default.
- W4281681567 hasConcept C2777211547 @default.
- W4281681567 hasConcept C2780586970 @default.
- W4281681567 hasConcept C41008148 @default.
- W4281681567 hasConcept C50644808 @default.
- W4281681567 hasConcept C51632099 @default.
- W4281681567 hasConcept C58973888 @default.
- W4281681567 hasConcept C77805123 @default.
- W4281681567 hasConcept C95623464 @default.
- W4281681567 hasConceptScore W4281681567C112972136 @default.
- W4281681567 hasConceptScore W4281681567C119857082 @default.
- W4281681567 hasConceptScore W4281681567C121332964 @default.
- W4281681567 hasConceptScore W4281681567C13280743 @default.
- W4281681567 hasConceptScore W4281681567C136389625 @default.
- W4281681567 hasConceptScore W4281681567C153180895 @default.
- W4281681567 hasConceptScore W4281681567C153294291 @default.
- W4281681567 hasConceptScore W4281681567C154945302 @default.
- W4281681567 hasConceptScore W4281681567C15744967 @default.
- W4281681567 hasConceptScore W4281681567C185798385 @default.
- W4281681567 hasConceptScore W4281681567C205649164 @default.
- W4281681567 hasConceptScore W4281681567C2776145971 @default.
- W4281681567 hasConceptScore W4281681567C2777211547 @default.
- W4281681567 hasConceptScore W4281681567C2780586970 @default.
- W4281681567 hasConceptScore W4281681567C41008148 @default.
- W4281681567 hasConceptScore W4281681567C50644808 @default.
- W4281681567 hasConceptScore W4281681567C51632099 @default.
- W4281681567 hasConceptScore W4281681567C58973888 @default.
- W4281681567 hasConceptScore W4281681567C77805123 @default.
- W4281681567 hasConceptScore W4281681567C95623464 @default.
- W4281681567 hasLocation W42816815671 @default.
- W4281681567 hasLocation W42816815672 @default.
- W4281681567 hasOpenAccess W4281681567 @default.
- W4281681567 hasPrimaryLocation W42816815671 @default.
- W4281681567 hasRelatedWork W1473009882 @default.
- W4281681567 hasRelatedWork W2104864891 @default.
- W4281681567 hasRelatedWork W2151561819 @default.
- W4281681567 hasRelatedWork W2538661024 @default.
- W4281681567 hasRelatedWork W4280650321 @default.
- W4281681567 hasRelatedWork W4294974824 @default.
- W4281681567 hasRelatedWork W4306904961 @default.
- W4281681567 hasRelatedWork W4312390859 @default.
- W4281681567 hasRelatedWork W4361193897 @default.
- W4281681567 hasRelatedWork W4385422125 @default.
- W4281681567 isParatext "false" @default.
- W4281681567 isRetracted "false" @default.
- W4281681567 workType "article" @default.