Matches in SemOpenAlex for { <https://semopenalex.org/work/W3132398984> ?p ?o ?g. }
- W3132398984 abstract "As machine learning techniques become widely adopted in new domains, especially in safety-critical systems such as autonomous vehicles, it is crucial to provide accurate output uncertainty estimation. As a result, many approaches have been proposed to calibrate neural networks to accurately estimate the likelihood of misclassification. However, while these methods achieve low calibration error, there is space for further improvement, especially in large-dimensional settings such as ImageNet. In this paper, we introduce a calibration algorithm, named Hoki, that works by applying random transformations to the neural network logits. We provide a sufficient condition for perfect calibration based on the number of label prediction changes observed after applying the transformations. We perform experiments on multiple datasets and show that the proposed approach generally outperforms state-of-the-art calibration algorithms across multiple datasets and models, especially on the challenging ImageNet dataset. Finally, Hoki is scalable as well, as it requires comparable execution time to that of temperature scaling." @default.
- W3132398984 created "2021-03-01" @default.
- W3132398984 creator A5030456600 @default.
- W3132398984 creator A5054356243 @default.
- W3132398984 creator A5060490339 @default.
- W3132398984 creator A5068946140 @default.
- W3132398984 date "2021-02-24" @default.
- W3132398984 modified "2023-10-17" @default.
- W3132398984 title "Confidence Calibration with Bounded Error Using Transformations" @default.
- W3132398984 cites W1618905105 @default.
- W3132398984 cites W1901129140 @default.
- W3132398984 cites W2108598243 @default.
- W3132398984 cites W2112796928 @default.
- W3132398984 cites W2194775991 @default.
- W3132398984 cites W2254249950 @default.
- W3132398984 cites W2331143823 @default.
- W3132398984 cites W2401231614 @default.
- W3132398984 cites W2612434969 @default.
- W3132398984 cites W2626967530 @default.
- W3132398984 cites W2796347433 @default.
- W3132398984 cites W2804697534 @default.
- W3132398984 cites W2933254221 @default.
- W3132398984 cites W2955425717 @default.
- W3132398984 cites W2963036823 @default.
- W3132398984 cites W2963163009 @default.
- W3132398984 cites W2963446712 @default.
- W3132398984 cites W2969338701 @default.
- W3132398984 cites W2971118045 @default.
- W3132398984 cites W2974098375 @default.
- W3132398984 cites W2974726644 @default.
- W3132398984 cites W2982509564 @default.
- W3132398984 cites W3018757597 @default.
- W3132398984 cites W3036157921 @default.
- W3132398984 cites W3037275730 @default.
- W3132398984 cites W3038321616 @default.
- W3132398984 cites W3095327551 @default.
- W3132398984 cites W3098898971 @default.
- W3132398984 cites W3105676370 @default.
- W3132398984 cites W3118608800 @default.
- W3132398984 cites W3132670545 @default.
- W3132398984 doi "https://doi.org/10.48550/arxiv.2102.12680" @default.
- W3132398984 hasPublicationYear "2021" @default.
- W3132398984 type Work @default.
- W3132398984 sameAs 3132398984 @default.
- W3132398984 citedByCount "0" @default.
- W3132398984 crossrefType "posted-content" @default.
- W3132398984 hasAuthorship W3132398984A5030456600 @default.
- W3132398984 hasAuthorship W3132398984A5054356243 @default.
- W3132398984 hasAuthorship W3132398984A5060490339 @default.
- W3132398984 hasAuthorship W3132398984A5068946140 @default.
- W3132398984 hasBestOaLocation W31323989841 @default.
- W3132398984 hasConcept C105795698 @default.
- W3132398984 hasConcept C11413529 @default.
- W3132398984 hasConcept C119857082 @default.
- W3132398984 hasConcept C124101348 @default.
- W3132398984 hasConcept C134306372 @default.
- W3132398984 hasConcept C154945302 @default.
- W3132398984 hasConcept C162324750 @default.
- W3132398984 hasConcept C165838908 @default.
- W3132398984 hasConcept C187736073 @default.
- W3132398984 hasConcept C2524010 @default.
- W3132398984 hasConcept C2780451532 @default.
- W3132398984 hasConcept C2781002164 @default.
- W3132398984 hasConcept C2984842247 @default.
- W3132398984 hasConcept C33923547 @default.
- W3132398984 hasConcept C34388435 @default.
- W3132398984 hasConcept C41008148 @default.
- W3132398984 hasConcept C48044578 @default.
- W3132398984 hasConcept C50644808 @default.
- W3132398984 hasConcept C77088390 @default.
- W3132398984 hasConcept C99844830 @default.
- W3132398984 hasConceptScore W3132398984C105795698 @default.
- W3132398984 hasConceptScore W3132398984C11413529 @default.
- W3132398984 hasConceptScore W3132398984C119857082 @default.
- W3132398984 hasConceptScore W3132398984C124101348 @default.
- W3132398984 hasConceptScore W3132398984C134306372 @default.
- W3132398984 hasConceptScore W3132398984C154945302 @default.
- W3132398984 hasConceptScore W3132398984C162324750 @default.
- W3132398984 hasConceptScore W3132398984C165838908 @default.
- W3132398984 hasConceptScore W3132398984C187736073 @default.
- W3132398984 hasConceptScore W3132398984C2524010 @default.
- W3132398984 hasConceptScore W3132398984C2780451532 @default.
- W3132398984 hasConceptScore W3132398984C2781002164 @default.
- W3132398984 hasConceptScore W3132398984C2984842247 @default.
- W3132398984 hasConceptScore W3132398984C33923547 @default.
- W3132398984 hasConceptScore W3132398984C34388435 @default.
- W3132398984 hasConceptScore W3132398984C41008148 @default.
- W3132398984 hasConceptScore W3132398984C48044578 @default.
- W3132398984 hasConceptScore W3132398984C50644808 @default.
- W3132398984 hasConceptScore W3132398984C77088390 @default.
- W3132398984 hasConceptScore W3132398984C99844830 @default.
- W3132398984 hasLocation W31323989841 @default.
- W3132398984 hasOpenAccess W3132398984 @default.
- W3132398984 hasPrimaryLocation W31323989841 @default.
- W3132398984 hasRelatedWork W1523472376 @default.
- W3132398984 hasRelatedWork W3020638616 @default.
- W3132398984 hasRelatedWork W3092824172 @default.
- W3132398984 hasRelatedWork W3105036711 @default.
- W3132398984 hasRelatedWork W3199608561 @default.
- W3132398984 hasRelatedWork W3200361725 @default.