Matches in SemOpenAlex for { <https://semopenalex.org/work/W3106195168> ?p ?o ?g. }
Showing items 1 to 76 of
76
with 100 items per page.
- W3106195168 endingPage "5491" @default.
- W3106195168 startingPage "5479" @default.
- W3106195168 abstract "Virtually any model we use in machine learning to make predictions does not perfectly represent reality. So, most of the learning happens under model misspecification. In this work, we present a novel analysis of the generalization performance of Bayesian model averaging under model misspecification and i.i.d. data using a new family of second-order PAC-Bayes bounds. This analysis shows, in simple and intuitive terms, that Bayesian model averaging provides suboptimal generalization performance when the model is misspecified. In consequence, we provide strong theoretical arguments showing that Bayesian methods are not optimal for learning predictive models, unless the model class is perfectly specified. Using novel second-order PAC-Bayes bounds, we derive a new family of Bayesian-like algorithms, which can be implemented as variational and ensemble methods. The output of these algorithms is a new posterior distribution, different from the Bayesian posterior, which induces a posterior predictive distribution with better generalization performance. Experiments with Bayesian neural networks illustrate these findings." @default.
- W3106195168 created "2020-11-23" @default.
- W3106195168 creator A5091366101 @default.
- W3106195168 date "2020-01-01" @default.
- W3106195168 modified "2023-10-18" @default.
- W3106195168 title "Learning under Model Misspecification: Applications to Variational and Ensemble methods" @default.
- W3106195168 hasPublicationYear "2020" @default.
- W3106195168 type Work @default.
- W3106195168 sameAs 3106195168 @default.
- W3106195168 citedByCount "14" @default.
- W3106195168 countsByYear W31061951682020 @default.
- W3106195168 countsByYear W31061951682021 @default.
- W3106195168 crossrefType "proceedings-article" @default.
- W3106195168 hasAuthorship W3106195168A5091366101 @default.
- W3106195168 hasConcept C107673813 @default.
- W3106195168 hasConcept C11413529 @default.
- W3106195168 hasConcept C119857082 @default.
- W3106195168 hasConcept C127043819 @default.
- W3106195168 hasConcept C134306372 @default.
- W3106195168 hasConcept C142291917 @default.
- W3106195168 hasConcept C149569020 @default.
- W3106195168 hasConcept C154945302 @default.
- W3106195168 hasConcept C160234255 @default.
- W3106195168 hasConcept C177148314 @default.
- W3106195168 hasConcept C207201462 @default.
- W3106195168 hasConcept C33923547 @default.
- W3106195168 hasConcept C41008148 @default.
- W3106195168 hasConcept C45942800 @default.
- W3106195168 hasConcept C57830394 @default.
- W3106195168 hasConcept C71983512 @default.
- W3106195168 hasConceptScore W3106195168C107673813 @default.
- W3106195168 hasConceptScore W3106195168C11413529 @default.
- W3106195168 hasConceptScore W3106195168C119857082 @default.
- W3106195168 hasConceptScore W3106195168C127043819 @default.
- W3106195168 hasConceptScore W3106195168C134306372 @default.
- W3106195168 hasConceptScore W3106195168C142291917 @default.
- W3106195168 hasConceptScore W3106195168C149569020 @default.
- W3106195168 hasConceptScore W3106195168C154945302 @default.
- W3106195168 hasConceptScore W3106195168C160234255 @default.
- W3106195168 hasConceptScore W3106195168C177148314 @default.
- W3106195168 hasConceptScore W3106195168C207201462 @default.
- W3106195168 hasConceptScore W3106195168C33923547 @default.
- W3106195168 hasConceptScore W3106195168C41008148 @default.
- W3106195168 hasConceptScore W3106195168C45942800 @default.
- W3106195168 hasConceptScore W3106195168C57830394 @default.
- W3106195168 hasConceptScore W3106195168C71983512 @default.
- W3106195168 hasLocation W31061951681 @default.
- W3106195168 hasOpenAccess W3106195168 @default.
- W3106195168 hasPrimaryLocation W31061951681 @default.
- W3106195168 hasRelatedWork W1788616518 @default.
- W3106195168 hasRelatedWork W2095652365 @default.
- W3106195168 hasRelatedWork W2194775991 @default.
- W3106195168 hasRelatedWork W2739217148 @default.
- W3106195168 hasRelatedWork W2752086744 @default.
- W3106195168 hasRelatedWork W2757830207 @default.
- W3106195168 hasRelatedWork W2785874076 @default.
- W3106195168 hasRelatedWork W2962894765 @default.
- W3106195168 hasRelatedWork W2963238274 @default.
- W3106195168 hasRelatedWork W2964059111 @default.
- W3106195168 hasRelatedWork W2964212410 @default.
- W3106195168 hasRelatedWork W2995924146 @default.
- W3106195168 hasRelatedWork W2996369242 @default.
- W3106195168 hasRelatedWork W3008152173 @default.
- W3106195168 hasRelatedWork W3034669169 @default.
- W3106195168 hasRelatedWork W3036446794 @default.
- W3106195168 hasRelatedWork W3101380508 @default.
- W3106195168 hasRelatedWork W3166517014 @default.
- W3106195168 hasRelatedWork W3201011253 @default.
- W3106195168 hasRelatedWork W599376124 @default.
- W3106195168 hasVolume "33" @default.
- W3106195168 isParatext "false" @default.
- W3106195168 isRetracted "false" @default.
- W3106195168 magId "3106195168" @default.
- W3106195168 workType "article" @default.