Matches in SemOpenAlex for { <https://semopenalex.org/work/W3037311587> ?p ?o ?g. }
- W3037311587 abstract "Recent success of large-scale pre-trained language models crucially hinge on fine-tuning them on large amounts of labeled data for the downstream task, that are typically expensive to acquire. In this work, we study self-training as one of the earliest semi-supervised learning approaches to reduce the annotation bottleneck by making use of large-scale unlabeled data for the target task. Standard self-training mechanism randomly samples instances from the unlabeled pool to pseudo-label and augment labeled data. In this work, we propose an approach to improve self-training by incorporating uncertainty estimates of the underlying neural network leveraging recent advances in Bayesian deep learning. Specifically, we propose (i) acquisition functions to select instances from the unlabeled pool leveraging Monte Carlo (MC) Dropout, and (ii) learning mechanism leveraging model confidence for self-training. As an application, we focus on text classification on five benchmark datasets. We show our methods leveraging only 20-30 labeled samples per class for each task for training and for validation can perform within 3% of fully supervised pre-trained language models fine-tuned on thousands of labeled instances with an aggregate accuracy of 91% and improving by upto 12% over baselines." @default.
- W3037311587 created "2020-07-02" @default.
- W3037311587 creator A5021000040 @default.
- W3037311587 creator A5033994052 @default.
- W3037311587 date "2020-06-27" @default.
- W3037311587 modified "2023-09-27" @default.
- W3037311587 title "Uncertainty-aware Self-training for Text Classification with Few Labels." @default.
- W3037311587 cites W1872312298 @default.
- W3037311587 cites W1995875735 @default.
- W3037311587 cites W2061873838 @default.
- W3037311587 cites W2095705004 @default.
- W3037311587 cites W2108501770 @default.
- W3037311587 cites W2108677974 @default.
- W3037311587 cites W2111316763 @default.
- W3037311587 cites W2113459411 @default.
- W3037311587 cites W2129068307 @default.
- W3037311587 cites W2132984949 @default.
- W3037311587 cites W2170973209 @default.
- W3037311587 cites W2251939518 @default.
- W3037311587 cites W2296073425 @default.
- W3037311587 cites W2526471240 @default.
- W3037311587 cites W2597787948 @default.
- W3037311587 cites W2740306440 @default.
- W3037311587 cites W2809669277 @default.
- W3037311587 cites W2897767292 @default.
- W3037311587 cites W2915348504 @default.
- W3037311587 cites W2962369866 @default.
- W3037311587 cites W2962973320 @default.
- W3037311587 cites W2963012544 @default.
- W3037311587 cites W2963207607 @default.
- W3037311587 cites W2963216553 @default.
- W3037311587 cites W2963341956 @default.
- W3037311587 cites W2963435192 @default.
- W3037311587 cites W2963476860 @default.
- W3037311587 cites W2963516811 @default.
- W3037311587 cites W2964059111 @default.
- W3037311587 cites W2964121744 @default.
- W3037311587 cites W2965373594 @default.
- W3037311587 cites W2970771703 @default.
- W3037311587 cites W2990940889 @default.
- W3037311587 cites W2995746049 @default.
- W3037311587 cites W3042491947 @default.
- W3037311587 cites W3136604105 @default.
- W3037311587 cites W3137695714 @default.
- W3037311587 cites W601603264 @default.
- W3037311587 cites W830076066 @default.
- W3037311587 cites W2530816535 @default.
- W3037311587 hasPublicationYear "2020" @default.
- W3037311587 type Work @default.
- W3037311587 sameAs 3037311587 @default.
- W3037311587 citedByCount "19" @default.
- W3037311587 countsByYear W30373115872020 @default.
- W3037311587 countsByYear W30373115872021 @default.
- W3037311587 crossrefType "posted-content" @default.
- W3037311587 hasAuthorship W3037311587A5021000040 @default.
- W3037311587 hasAuthorship W3037311587A5033994052 @default.
- W3037311587 hasConcept C107673813 @default.
- W3037311587 hasConcept C119857082 @default.
- W3037311587 hasConcept C13280743 @default.
- W3037311587 hasConcept C149635348 @default.
- W3037311587 hasConcept C154945302 @default.
- W3037311587 hasConcept C162324750 @default.
- W3037311587 hasConcept C185798385 @default.
- W3037311587 hasConcept C187736073 @default.
- W3037311587 hasConcept C205649164 @default.
- W3037311587 hasConcept C2776145597 @default.
- W3037311587 hasConcept C2776145971 @default.
- W3037311587 hasConcept C2776321320 @default.
- W3037311587 hasConcept C2780451532 @default.
- W3037311587 hasConcept C2780513914 @default.
- W3037311587 hasConcept C41008148 @default.
- W3037311587 hasConceptScore W3037311587C107673813 @default.
- W3037311587 hasConceptScore W3037311587C119857082 @default.
- W3037311587 hasConceptScore W3037311587C13280743 @default.
- W3037311587 hasConceptScore W3037311587C149635348 @default.
- W3037311587 hasConceptScore W3037311587C154945302 @default.
- W3037311587 hasConceptScore W3037311587C162324750 @default.
- W3037311587 hasConceptScore W3037311587C185798385 @default.
- W3037311587 hasConceptScore W3037311587C187736073 @default.
- W3037311587 hasConceptScore W3037311587C205649164 @default.
- W3037311587 hasConceptScore W3037311587C2776145597 @default.
- W3037311587 hasConceptScore W3037311587C2776145971 @default.
- W3037311587 hasConceptScore W3037311587C2776321320 @default.
- W3037311587 hasConceptScore W3037311587C2780451532 @default.
- W3037311587 hasConceptScore W3037311587C2780513914 @default.
- W3037311587 hasConceptScore W3037311587C41008148 @default.
- W3037311587 hasLocation W30373115871 @default.
- W3037311587 hasOpenAccess W3037311587 @default.
- W3037311587 hasPrimaryLocation W30373115871 @default.
- W3037311587 hasRelatedWork W2111316763 @default.
- W3037311587 hasRelatedWork W2902917913 @default.
- W3037311587 hasRelatedWork W2911489562 @default.
- W3037311587 hasRelatedWork W2914913933 @default.
- W3037311587 hasRelatedWork W2949092679 @default.
- W3037311587 hasRelatedWork W2962739339 @default.
- W3037311587 hasRelatedWork W2963341956 @default.
- W3037311587 hasRelatedWork W2964005754 @default.
- W3037311587 hasRelatedWork W2965373594 @default.
- W3037311587 hasRelatedWork W2981571980 @default.
- W3037311587 hasRelatedWork W2995746049 @default.