Matches in SemOpenAlex for { <https://semopenalex.org/work/W3206219953> ?p ?o ?g. }
- W3206219953 abstract "We currently do not have an understanding of semi-supervised learning (SSL) objectives such as pseudo-labelling and entropy minimization as log-likelihoods, which precludes the development of e.g. Bayesian SSL. Here, we note that benchmark image datasets such as CIFAR-10 are carefully curated, and we formulate SSL objectives as a log-likelihood in a generative model of data curation that was initially developed to explain the cold-posterior effect (Aitchison 2020). SSL methods, from entropy minimization and pseudo-labelling, to state-of-the-art techniques similar to FixMatch can be understood as lower-bounds on our principled log-likelihood. We are thus able to give a proof-of-principle for Bayesian SSL on toy data. Finally, our theory suggests that SSL is effective in part due to the statistical patterns induced by data curation. This provides an explanation of past results which show SSL performs better on clean datasets without any out of distribution examples. Confirming these results we find that SSL gave much larger performance improvements on curated than on uncurated data, using matched curated and uncurated datasets based on Galaxy Zoo 2." @default.
- W3206219953 created "2021-10-25" @default.
- W3206219953 creator A5057390894 @default.
- W3206219953 creator A5062539506 @default.
- W3206219953 date "2021-10-10" @default.
- W3206219953 modified "2023-09-26" @default.
- W3206219953 title "Semi-supervised learning objectives as log-likelihoods in a generative model of data curation." @default.
- W3206219953 cites W1532801683 @default.
- W3206219953 cites W1574877594 @default.
- W3206219953 cites W1630959083 @default.
- W3206219953 cites W1975165783 @default.
- W3206219953 cites W1990334093 @default.
- W3206219953 cites W2019288156 @default.
- W3206219953 cites W2023037133 @default.
- W3206219953 cites W2083515729 @default.
- W3206219953 cites W2089053726 @default.
- W3206219953 cites W2104755048 @default.
- W3206219953 cites W2108501770 @default.
- W3206219953 cites W2108598243 @default.
- W3206219953 cites W2136504847 @default.
- W3206219953 cites W2145366000 @default.
- W3206219953 cites W2145494108 @default.
- W3206219953 cites W2148603752 @default.
- W3206219953 cites W2412510955 @default.
- W3206219953 cites W2724955321 @default.
- W3206219953 cites W2753738274 @default.
- W3206219953 cites W2892099372 @default.
- W3206219953 cites W2962369866 @default.
- W3206219953 cites W2963435192 @default.
- W3206219953 cites W2963956526 @default.
- W3206219953 cites W2964051842 @default.
- W3206219953 cites W2964159205 @default.
- W3206219953 cites W2978426779 @default.
- W3206219953 cites W2984353870 @default.
- W3206219953 cites W2996501936 @default.
- W3206219953 cites W2997131443 @default.
- W3206219953 cites W3001197829 @default.
- W3206219953 cites W3034669169 @default.
- W3206219953 cites W3118608800 @default.
- W3206219953 cites W3123387470 @default.
- W3206219953 cites W3140968660 @default.
- W3206219953 hasPublicationYear "2021" @default.
- W3206219953 type Work @default.
- W3206219953 sameAs 3206219953 @default.
- W3206219953 citedByCount "0" @default.
- W3206219953 crossrefType "posted-content" @default.
- W3206219953 hasAuthorship W3206219953A5057390894 @default.
- W3206219953 hasAuthorship W3206219953A5062539506 @default.
- W3206219953 hasConcept C106301342 @default.
- W3206219953 hasConcept C107673813 @default.
- W3206219953 hasConcept C119857082 @default.
- W3206219953 hasConcept C121332964 @default.
- W3206219953 hasConcept C13280743 @default.
- W3206219953 hasConcept C154945302 @default.
- W3206219953 hasConcept C167966045 @default.
- W3206219953 hasConcept C185798385 @default.
- W3206219953 hasConcept C205649164 @default.
- W3206219953 hasConcept C39890363 @default.
- W3206219953 hasConcept C41008148 @default.
- W3206219953 hasConcept C62520636 @default.
- W3206219953 hasConcept C9679016 @default.
- W3206219953 hasConceptScore W3206219953C106301342 @default.
- W3206219953 hasConceptScore W3206219953C107673813 @default.
- W3206219953 hasConceptScore W3206219953C119857082 @default.
- W3206219953 hasConceptScore W3206219953C121332964 @default.
- W3206219953 hasConceptScore W3206219953C13280743 @default.
- W3206219953 hasConceptScore W3206219953C154945302 @default.
- W3206219953 hasConceptScore W3206219953C167966045 @default.
- W3206219953 hasConceptScore W3206219953C185798385 @default.
- W3206219953 hasConceptScore W3206219953C205649164 @default.
- W3206219953 hasConceptScore W3206219953C39890363 @default.
- W3206219953 hasConceptScore W3206219953C41008148 @default.
- W3206219953 hasConceptScore W3206219953C62520636 @default.
- W3206219953 hasConceptScore W3206219953C9679016 @default.
- W3206219953 hasLocation W32062199531 @default.
- W3206219953 hasOpenAccess W3206219953 @default.
- W3206219953 hasPrimaryLocation W32062199531 @default.
- W3206219953 hasRelatedWork W1585902826 @default.
- W3206219953 hasRelatedWork W2017318051 @default.
- W3206219953 hasRelatedWork W2109322901 @default.
- W3206219953 hasRelatedWork W2608412030 @default.
- W3206219953 hasRelatedWork W2752794725 @default.
- W3206219953 hasRelatedWork W2914569968 @default.
- W3206219953 hasRelatedWork W2947188859 @default.
- W3206219953 hasRelatedWork W2999527990 @default.
- W3206219953 hasRelatedWork W3001197829 @default.
- W3206219953 hasRelatedWork W3013561810 @default.
- W3206219953 hasRelatedWork W3030172965 @default.
- W3206219953 hasRelatedWork W3039156018 @default.
- W3206219953 hasRelatedWork W3108067698 @default.
- W3206219953 hasRelatedWork W3119385233 @default.
- W3206219953 hasRelatedWork W3200852773 @default.
- W3206219953 hasRelatedWork W3201315369 @default.
- W3206219953 hasRelatedWork W3202133550 @default.
- W3206219953 hasRelatedWork W3203521893 @default.
- W3206219953 hasRelatedWork W3204006979 @default.
- W3206219953 hasRelatedWork W3211170372 @default.
- W3206219953 isParatext "false" @default.
- W3206219953 isRetracted "false" @default.
- W3206219953 magId "3206219953" @default.