Matches in SemOpenAlex for { <https://semopenalex.org/work/W4378572066> ?p ?o ?g. }
Showing items 1 to 59 of
59
with 100 items per page.
- W4378572066 abstract "Current algorithms and architecture can create excellent DNN classifier models from example data. In general, larger training datasets result in better model estimations, which improve test performance. Existing methods for predicting generalization performance are based on hold-out test examples. To the best of our knowledge, at present no method exists that can estimate the quality of a trained DNN classifier without test data. In this paper, we show that the quality of a trained DNN classifier can be assessed without any example data. We consider DNNs to be composed of a feature extractor and a feature classifier; the feature extractor's output is fed to the classifier. The proposed method iteratively creates class prototypes in the input space for each class by minimizing a cross-entropy loss function at the output of the network. We use these prototypes and their feature relationships to reveal the quality of the classifier. We have developed two metrics: one using the features of the prototypes and the other using adversarial examples corresponding to each prototype. Empirical evaluations show that accuracy obtained from test examples is directly proportional to quality measures obtained from the proposed metrics. We report our observations for ResNet18 with Tiny ImageNet, CIFAR100, and CIFAR10 datasets. The proposed metrics can be used to compare performances of two or more classifiers without test examples." @default.
- W4378572066 created "2023-05-28" @default.
- W4378572066 creator A5004229694 @default.
- W4378572066 creator A5072973533 @default.
- W4378572066 date "2023-05-24" @default.
- W4378572066 modified "2023-09-26" @default.
- W4378572066 title "Fantastic DNN Classifiers and How to Identify them without Data" @default.
- W4378572066 doi "https://doi.org/10.48550/arxiv.2305.15563" @default.
- W4378572066 hasPublicationYear "2023" @default.
- W4378572066 type Work @default.
- W4378572066 citedByCount "0" @default.
- W4378572066 crossrefType "posted-content" @default.
- W4378572066 hasAuthorship W4378572066A5004229694 @default.
- W4378572066 hasAuthorship W4378572066A5072973533 @default.
- W4378572066 hasBestOaLocation W43785720661 @default.
- W4378572066 hasConcept C117978034 @default.
- W4378572066 hasConcept C119857082 @default.
- W4378572066 hasConcept C124101348 @default.
- W4378572066 hasConcept C127413603 @default.
- W4378572066 hasConcept C153180895 @default.
- W4378572066 hasConcept C154945302 @default.
- W4378572066 hasConcept C16910744 @default.
- W4378572066 hasConcept C173102733 @default.
- W4378572066 hasConcept C199360897 @default.
- W4378572066 hasConcept C21880701 @default.
- W4378572066 hasConcept C41008148 @default.
- W4378572066 hasConcept C50644808 @default.
- W4378572066 hasConcept C83665646 @default.
- W4378572066 hasConcept C95623464 @default.
- W4378572066 hasConceptScore W4378572066C117978034 @default.
- W4378572066 hasConceptScore W4378572066C119857082 @default.
- W4378572066 hasConceptScore W4378572066C124101348 @default.
- W4378572066 hasConceptScore W4378572066C127413603 @default.
- W4378572066 hasConceptScore W4378572066C153180895 @default.
- W4378572066 hasConceptScore W4378572066C154945302 @default.
- W4378572066 hasConceptScore W4378572066C16910744 @default.
- W4378572066 hasConceptScore W4378572066C173102733 @default.
- W4378572066 hasConceptScore W4378572066C199360897 @default.
- W4378572066 hasConceptScore W4378572066C21880701 @default.
- W4378572066 hasConceptScore W4378572066C41008148 @default.
- W4378572066 hasConceptScore W4378572066C50644808 @default.
- W4378572066 hasConceptScore W4378572066C83665646 @default.
- W4378572066 hasConceptScore W4378572066C95623464 @default.
- W4378572066 hasLocation W43785720661 @default.
- W4378572066 hasOpenAccess W4378572066 @default.
- W4378572066 hasPrimaryLocation W43785720661 @default.
- W4378572066 hasRelatedWork W2062316447 @default.
- W4378572066 hasRelatedWork W2143407407 @default.
- W4378572066 hasRelatedWork W2160451891 @default.
- W4378572066 hasRelatedWork W2275058042 @default.
- W4378572066 hasRelatedWork W2371712871 @default.
- W4378572066 hasRelatedWork W2556319748 @default.
- W4378572066 hasRelatedWork W2765332597 @default.
- W4378572066 hasRelatedWork W4378572066 @default.
- W4378572066 hasRelatedWork W1822914510 @default.
- W4378572066 hasRelatedWork W2187500075 @default.
- W4378572066 isParatext "false" @default.
- W4378572066 isRetracted "false" @default.
- W4378572066 workType "article" @default.