Matches in SemOpenAlex for { <https://semopenalex.org/work/W3205001720> ?p ?o ?g. }
- W3205001720 abstract "MC Dropout is a mainstream free lunch method in medical imaging for approximate Bayesian computations (ABC). Its appeal is to solve out-of-the-box the daunting task of ABC and uncertainty quantification in Neural Networks (NNs); to fall within the variational inference (VI) framework; and to propose a highly multimodal, faithful predictive posterior. We question the properties of MC Dropout for approximate inference, as in fact MC Dropout changes the Bayesian model; its predictive posterior assigns $0$ probability to the true model on closed-form benchmarks; the multimodality of its predictive posterior is not a property of the true predictive posterior but a design artefact. To address the need for VI on arbitrary models, we share a generic VI engine within the pytorch framework. The code includes a carefully designed implementation of structured (diagonal plus low-rank) multivariate normal variational families, and mixtures thereof. It is intended as a go-to no-free-lunch approach, addressing shortcomings of mean-field VI with an adjustable trade-off between expressivity and computational complexity." @default.
- W3205001720 created "2021-10-25" @default.
- W3205001720 creator A5007222325 @default.
- W3205001720 creator A5019012882 @default.
- W3205001720 creator A5033246044 @default.
- W3205001720 creator A5037306463 @default.
- W3205001720 creator A5038438003 @default.
- W3205001720 creator A5056261326 @default.
- W3205001720 creator A5063499796 @default.
- W3205001720 creator A5070704852 @default.
- W3205001720 creator A5077626501 @default.
- W3205001720 creator A5085577709 @default.
- W3205001720 date "2021-10-08" @default.
- W3205001720 modified "2023-09-27" @default.
- W3205001720 title "Is MC Dropout Bayesian" @default.
- W3205001720 cites W1663973292 @default.
- W3205001720 cites W1973333099 @default.
- W3205001720 cites W2082030398 @default.
- W3205001720 cites W2089940272 @default.
- W3205001720 cites W2095705004 @default.
- W3205001720 cites W2096104495 @default.
- W3205001720 cites W2104465614 @default.
- W3205001720 cites W2161455851 @default.
- W3205001720 cites W2548353158 @default.
- W3205001720 cites W2552233335 @default.
- W3205001720 cites W2559930736 @default.
- W3205001720 cites W2592505114 @default.
- W3205001720 cites W2592796789 @default.
- W3205001720 cites W2594374979 @default.
- W3205001720 cites W2731982447 @default.
- W3205001720 cites W2752620087 @default.
- W3205001720 cites W2785391921 @default.
- W3205001720 cites W2794396789 @default.
- W3205001720 cites W2803806607 @default.
- W3205001720 cites W2806584432 @default.
- W3205001720 cites W2808208825 @default.
- W3205001720 cites W2951266961 @default.
- W3205001720 cites W2952240579 @default.
- W3205001720 cites W2962851448 @default.
- W3205001720 cites W2963153890 @default.
- W3205001720 cites W2964059111 @default.
- W3205001720 hasPublicationYear "2021" @default.
- W3205001720 type Work @default.
- W3205001720 sameAs 3205001720 @default.
- W3205001720 citedByCount "0" @default.
- W3205001720 crossrefType "posted-content" @default.
- W3205001720 hasAuthorship W3205001720A5007222325 @default.
- W3205001720 hasAuthorship W3205001720A5019012882 @default.
- W3205001720 hasAuthorship W3205001720A5033246044 @default.
- W3205001720 hasAuthorship W3205001720A5037306463 @default.
- W3205001720 hasAuthorship W3205001720A5038438003 @default.
- W3205001720 hasAuthorship W3205001720A5056261326 @default.
- W3205001720 hasAuthorship W3205001720A5063499796 @default.
- W3205001720 hasAuthorship W3205001720A5070704852 @default.
- W3205001720 hasAuthorship W3205001720A5077626501 @default.
- W3205001720 hasAuthorship W3205001720A5085577709 @default.
- W3205001720 hasConcept C107673813 @default.
- W3205001720 hasConcept C11413529 @default.
- W3205001720 hasConcept C119857082 @default.
- W3205001720 hasConcept C154945302 @default.
- W3205001720 hasConcept C160234255 @default.
- W3205001720 hasConcept C207201462 @default.
- W3205001720 hasConcept C2776145597 @default.
- W3205001720 hasConcept C2776214188 @default.
- W3205001720 hasConcept C41008148 @default.
- W3205001720 hasConcept C57830394 @default.
- W3205001720 hasConceptScore W3205001720C107673813 @default.
- W3205001720 hasConceptScore W3205001720C11413529 @default.
- W3205001720 hasConceptScore W3205001720C119857082 @default.
- W3205001720 hasConceptScore W3205001720C154945302 @default.
- W3205001720 hasConceptScore W3205001720C160234255 @default.
- W3205001720 hasConceptScore W3205001720C207201462 @default.
- W3205001720 hasConceptScore W3205001720C2776145597 @default.
- W3205001720 hasConceptScore W3205001720C2776214188 @default.
- W3205001720 hasConceptScore W3205001720C41008148 @default.
- W3205001720 hasConceptScore W3205001720C57830394 @default.
- W3205001720 hasLocation W32050017201 @default.
- W3205001720 hasOpenAccess W3205001720 @default.
- W3205001720 hasPrimaryLocation W32050017201 @default.
- W3205001720 hasRelatedWork W1971038356 @default.
- W3205001720 hasRelatedWork W1971722701 @default.
- W3205001720 hasRelatedWork W2137123230 @default.
- W3205001720 hasRelatedWork W2155593578 @default.
- W3205001720 hasRelatedWork W2187318208 @default.
- W3205001720 hasRelatedWork W2885229895 @default.
- W3205001720 hasRelatedWork W2910171846 @default.
- W3205001720 hasRelatedWork W2920569681 @default.
- W3205001720 hasRelatedWork W2947972834 @default.
- W3205001720 hasRelatedWork W2952264928 @default.
- W3205001720 hasRelatedWork W2962775468 @default.
- W3205001720 hasRelatedWork W2963545688 @default.
- W3205001720 hasRelatedWork W3005677713 @default.
- W3205001720 hasRelatedWork W3008064687 @default.
- W3205001720 hasRelatedWork W3033729636 @default.
- W3205001720 hasRelatedWork W3093963773 @default.
- W3205001720 hasRelatedWork W3131926344 @default.
- W3205001720 hasRelatedWork W3132714944 @default.
- W3205001720 hasRelatedWork W3176960246 @default.
- W3205001720 hasRelatedWork W62820921 @default.
- W3205001720 isParatext "false" @default.