Matches in SemOpenAlex for { <https://semopenalex.org/work/W3213016052> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W3213016052 abstract "Deep Learning methods are known to suffer from calibration issues: they typically produce over-confident estimates. These problems are exacerbated in the low data regime. Although the calibration of probabilistic models is well studied, calibrating extremely over-parametrized models in the low-data regime presents unique challenges. We show that deep-ensembles do not necessarily lead to improved calibration properties. In fact, we show that standard ensembling methods, when used in conjunction with modern techniques such as mixup regularization, can lead to less calibrated models. In this text, we examine the interplay between three of the most simple and commonly used approaches to leverage deep learning when data is scarce: data-augmentation, ensembling, and post-processing calibration methods. We demonstrate that, although standard ensembling techniques certainly help to boost accuracy, the calibration of deep-ensembles relies on subtle trade-offs. Our main finding is that calibration methods such as temperature scaling need to be slightly tweaked when used with deep-ensembles and, crucially, need to be executed after the averaging process. Our simulations indicate that, in the low data regime, this simple strategy can halve the Expected Calibration Error (ECE) on a range of benchmark classification problems when compared to standard deep-ensembles." @default.
- W3213016052 created "2021-11-22" @default.
- W3213016052 creator A5060952088 @default.
- W3213016052 creator A5075419354 @default.
- W3213016052 date "2021-12-06" @default.
- W3213016052 modified "2023-09-26" @default.
- W3213016052 title "Uncertainty Quantification and Deep Ensembles" @default.
- W3213016052 hasPublicationYear "2021" @default.
- W3213016052 type Work @default.
- W3213016052 sameAs 3213016052 @default.
- W3213016052 citedByCount "0" @default.
- W3213016052 crossrefType "proceedings-article" @default.
- W3213016052 hasAuthorship W3213016052A5060952088 @default.
- W3213016052 hasAuthorship W3213016052A5075419354 @default.
- W3213016052 hasConcept C105795698 @default.
- W3213016052 hasConcept C108583219 @default.
- W3213016052 hasConcept C119857082 @default.
- W3213016052 hasConcept C124101348 @default.
- W3213016052 hasConcept C13280743 @default.
- W3213016052 hasConcept C153083717 @default.
- W3213016052 hasConcept C154945302 @default.
- W3213016052 hasConcept C159985019 @default.
- W3213016052 hasConcept C165838908 @default.
- W3213016052 hasConcept C185798385 @default.
- W3213016052 hasConcept C192562407 @default.
- W3213016052 hasConcept C204323151 @default.
- W3213016052 hasConcept C205649164 @default.
- W3213016052 hasConcept C2776135515 @default.
- W3213016052 hasConcept C2984842247 @default.
- W3213016052 hasConcept C33923547 @default.
- W3213016052 hasConcept C41008148 @default.
- W3213016052 hasConcept C49937458 @default.
- W3213016052 hasConceptScore W3213016052C105795698 @default.
- W3213016052 hasConceptScore W3213016052C108583219 @default.
- W3213016052 hasConceptScore W3213016052C119857082 @default.
- W3213016052 hasConceptScore W3213016052C124101348 @default.
- W3213016052 hasConceptScore W3213016052C13280743 @default.
- W3213016052 hasConceptScore W3213016052C153083717 @default.
- W3213016052 hasConceptScore W3213016052C154945302 @default.
- W3213016052 hasConceptScore W3213016052C159985019 @default.
- W3213016052 hasConceptScore W3213016052C165838908 @default.
- W3213016052 hasConceptScore W3213016052C185798385 @default.
- W3213016052 hasConceptScore W3213016052C192562407 @default.
- W3213016052 hasConceptScore W3213016052C204323151 @default.
- W3213016052 hasConceptScore W3213016052C205649164 @default.
- W3213016052 hasConceptScore W3213016052C2776135515 @default.
- W3213016052 hasConceptScore W3213016052C2984842247 @default.
- W3213016052 hasConceptScore W3213016052C33923547 @default.
- W3213016052 hasConceptScore W3213016052C41008148 @default.
- W3213016052 hasConceptScore W3213016052C49937458 @default.
- W3213016052 hasLocation W32130160521 @default.
- W3213016052 hasOpenAccess W3213016052 @default.
- W3213016052 hasPrimaryLocation W32130160521 @default.
- W3213016052 hasRelatedWork W2788453699 @default.
- W3213016052 hasRelatedWork W2948194985 @default.
- W3213016052 hasRelatedWork W2950896278 @default.
- W3213016052 hasRelatedWork W3003830418 @default.
- W3213016052 hasRelatedWork W3007458866 @default.
- W3213016052 hasRelatedWork W3010624034 @default.
- W3213016052 hasRelatedWork W3012855981 @default.
- W3213016052 hasRelatedWork W3017120154 @default.
- W3213016052 hasRelatedWork W3043321433 @default.
- W3213016052 hasRelatedWork W3094557955 @default.
- W3213016052 hasRelatedWork W3097920418 @default.
- W3213016052 hasRelatedWork W3104668038 @default.
- W3213016052 hasRelatedWork W3105518655 @default.
- W3213016052 hasRelatedWork W3112354295 @default.
- W3213016052 hasRelatedWork W3131643948 @default.
- W3213016052 hasRelatedWork W3132297750 @default.
- W3213016052 hasRelatedWork W3154139798 @default.
- W3213016052 hasRelatedWork W3156027305 @default.
- W3213016052 hasRelatedWork W3210898911 @default.
- W3213016052 hasRelatedWork W3214019887 @default.
- W3213016052 hasVolume "34" @default.
- W3213016052 isParatext "false" @default.
- W3213016052 isRetracted "false" @default.
- W3213016052 magId "3213016052" @default.
- W3213016052 workType "article" @default.