Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287316056> ?p ?o ?g. }
Showing items 1 to 59 of
59
with 100 items per page.
- W4287316056 abstract "Multi-label audio tagging consists of assigning sets of tags to audio recordings. At inference time, thresholds are applied on the confidence scores outputted by a probabilistic classifier, in order to decide which classes are detected active. In this work, we consider having at disposal a trained classifier and we seek to automatically optimize the decision thresholds according to a performance metric of interest, in our case F-measure (micro-F1). We propose a new method, called SGL-Thresh for Surrogate Gradient Learning of Thresholds, that makes use of gradient descent. Since F1 is not differentiable, we propose to approximate the thresholding operation gradients with the gradients of a sigmoid function. We report experiments on three datasets, using state-of-the-art pre-trained deep neural networks. In all cases, SGL-Thresh outperformed three other approaches: a default threshold value (defThresh), an heuristic search algorithm and a method estimating F1 gradients numerically. It reached 54.9% F1 on AudioSet eval, compared to 50.7% with defThresh. SGL-Thresh is very fast and scalable to a large number of tags. To facilitate reproducibility, data and source code in Pytorch are available online: https://github.com/topel/SGL-Thresh" @default.
- W4287316056 created "2022-07-25" @default.
- W4287316056 creator A5005884869 @default.
- W4287316056 creator A5088037163 @default.
- W4287316056 date "2021-03-01" @default.
- W4287316056 modified "2023-10-08" @default.
- W4287316056 title "Fast threshold optimization for multi-label audio tagging using Surrogate gradient learning" @default.
- W4287316056 doi "https://doi.org/10.48550/arxiv.2103.00833" @default.
- W4287316056 hasPublicationYear "2021" @default.
- W4287316056 type Work @default.
- W4287316056 citedByCount "0" @default.
- W4287316056 crossrefType "posted-content" @default.
- W4287316056 hasAuthorship W4287316056A5005884869 @default.
- W4287316056 hasAuthorship W4287316056A5088037163 @default.
- W4287316056 hasBestOaLocation W42873160561 @default.
- W4287316056 hasConcept C115961682 @default.
- W4287316056 hasConcept C119857082 @default.
- W4287316056 hasConcept C153180895 @default.
- W4287316056 hasConcept C153258448 @default.
- W4287316056 hasConcept C154945302 @default.
- W4287316056 hasConcept C191178318 @default.
- W4287316056 hasConcept C2776214188 @default.
- W4287316056 hasConcept C41008148 @default.
- W4287316056 hasConcept C48044578 @default.
- W4287316056 hasConcept C49937458 @default.
- W4287316056 hasConcept C50644808 @default.
- W4287316056 hasConcept C77088390 @default.
- W4287316056 hasConcept C95623464 @default.
- W4287316056 hasConceptScore W4287316056C115961682 @default.
- W4287316056 hasConceptScore W4287316056C119857082 @default.
- W4287316056 hasConceptScore W4287316056C153180895 @default.
- W4287316056 hasConceptScore W4287316056C153258448 @default.
- W4287316056 hasConceptScore W4287316056C154945302 @default.
- W4287316056 hasConceptScore W4287316056C191178318 @default.
- W4287316056 hasConceptScore W4287316056C2776214188 @default.
- W4287316056 hasConceptScore W4287316056C41008148 @default.
- W4287316056 hasConceptScore W4287316056C48044578 @default.
- W4287316056 hasConceptScore W4287316056C49937458 @default.
- W4287316056 hasConceptScore W4287316056C50644808 @default.
- W4287316056 hasConceptScore W4287316056C77088390 @default.
- W4287316056 hasConceptScore W4287316056C95623464 @default.
- W4287316056 hasLocation W42873160561 @default.
- W4287316056 hasLocation W42873160562 @default.
- W4287316056 hasLocation W42873160563 @default.
- W4287316056 hasOpenAccess W4287316056 @default.
- W4287316056 hasPrimaryLocation W42873160561 @default.
- W4287316056 hasRelatedWork W2033000528 @default.
- W4287316056 hasRelatedWork W2160069326 @default.
- W4287316056 hasRelatedWork W2563096758 @default.
- W4287316056 hasRelatedWork W2792520941 @default.
- W4287316056 hasRelatedWork W2961085424 @default.
- W4287316056 hasRelatedWork W3136652852 @default.
- W4287316056 hasRelatedWork W3175271736 @default.
- W4287316056 hasRelatedWork W4200631632 @default.
- W4287316056 hasRelatedWork W4226100381 @default.
- W4287316056 hasRelatedWork W51958864 @default.
- W4287316056 isParatext "false" @default.
- W4287316056 isRetracted "false" @default.
- W4287316056 workType "article" @default.