Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287555692> ?p ?o ?g. }
Showing items 1 to 59 of
59
with 100 items per page.
- W4287555692 abstract "We propose a kernelized classification layer for deep networks. Although conventional deep networks introduce an abundance of nonlinearity for representation (feature) learning, they almost universally use a linear classifier on the learned feature vectors. We advocate a nonlinear classification layer by using the kernel trick on the softmax cross-entropy loss function during training and the scorer function during testing. However, the choice of the kernel remains a challenge. To tackle this, we theoretically show the possibility of optimizing over all possible positive definite kernels applicable to our problem setting. This theory is then used to device a new kernelized classification layer that learns the optimal kernel function for a given problem automatically within the deep network itself. We show the usefulness of the proposed nonlinear classification layer on several datasets and tasks." @default.
- W4287555692 created "2022-07-25" @default.
- W4287555692 creator A5018301052 @default.
- W4287555692 creator A5060350100 @default.
- W4287555692 creator A5075594276 @default.
- W4287555692 date "2020-12-08" @default.
- W4287555692 modified "2023-10-17" @default.
- W4287555692 title "Kernelized Classification in Deep Networks" @default.
- W4287555692 doi "https://doi.org/10.48550/arxiv.2012.09607" @default.
- W4287555692 hasPublicationYear "2020" @default.
- W4287555692 type Work @default.
- W4287555692 citedByCount "0" @default.
- W4287555692 crossrefType "posted-content" @default.
- W4287555692 hasAuthorship W4287555692A5018301052 @default.
- W4287555692 hasAuthorship W4287555692A5060350100 @default.
- W4287555692 hasAuthorship W4287555692A5075594276 @default.
- W4287555692 hasBestOaLocation W42875556921 @default.
- W4287555692 hasConcept C108583219 @default.
- W4287555692 hasConcept C114614502 @default.
- W4287555692 hasConcept C119857082 @default.
- W4287555692 hasConcept C122280245 @default.
- W4287555692 hasConcept C12267149 @default.
- W4287555692 hasConcept C139532973 @default.
- W4287555692 hasConcept C153180895 @default.
- W4287555692 hasConcept C154945302 @default.
- W4287555692 hasConcept C188441871 @default.
- W4287555692 hasConcept C33923547 @default.
- W4287555692 hasConcept C41008148 @default.
- W4287555692 hasConcept C74193536 @default.
- W4287555692 hasConcept C95623464 @default.
- W4287555692 hasConceptScore W4287555692C108583219 @default.
- W4287555692 hasConceptScore W4287555692C114614502 @default.
- W4287555692 hasConceptScore W4287555692C119857082 @default.
- W4287555692 hasConceptScore W4287555692C122280245 @default.
- W4287555692 hasConceptScore W4287555692C12267149 @default.
- W4287555692 hasConceptScore W4287555692C139532973 @default.
- W4287555692 hasConceptScore W4287555692C153180895 @default.
- W4287555692 hasConceptScore W4287555692C154945302 @default.
- W4287555692 hasConceptScore W4287555692C188441871 @default.
- W4287555692 hasConceptScore W4287555692C33923547 @default.
- W4287555692 hasConceptScore W4287555692C41008148 @default.
- W4287555692 hasConceptScore W4287555692C74193536 @default.
- W4287555692 hasConceptScore W4287555692C95623464 @default.
- W4287555692 hasLocation W42875556921 @default.
- W4287555692 hasOpenAccess W4287555692 @default.
- W4287555692 hasPrimaryLocation W42875556921 @default.
- W4287555692 hasRelatedWork W2141705618 @default.
- W4287555692 hasRelatedWork W2160451891 @default.
- W4287555692 hasRelatedWork W2348964713 @default.
- W4287555692 hasRelatedWork W2539163683 @default.
- W4287555692 hasRelatedWork W2771515600 @default.
- W4287555692 hasRelatedWork W2803710604 @default.
- W4287555692 hasRelatedWork W2807311372 @default.
- W4287555692 hasRelatedWork W2971416272 @default.
- W4287555692 hasRelatedWork W3214058074 @default.
- W4287555692 hasRelatedWork W4221015625 @default.
- W4287555692 isParatext "false" @default.
- W4287555692 isRetracted "false" @default.
- W4287555692 workType "article" @default.