FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3036700096> ?p ?o ?g. }

• W3036700096 abstract "We study deep neural networks (DNNs) trained on natural image data with entirely random labels. Despite its popularity in the literature, where it is often used to study memorization, generalization, and other phenomena, little is known about what DNNs learn in this setting. In this paper, we show analytically for convolutional and fully connected networks that an alignment between the principal components of network parameters and data takes place when training with random labels. We study this alignment effect by investigating neural networks pre-trained on randomly labelled image data and subsequently fine-tuned on disjoint datasets with random or real labels. We show how this alignment produces a positive transfer: networks pre-trained with random labels train faster downstream compared to training from scratch even after accounting for simple effects, such as weight scaling. We analyze how competing effects, such as specialization at later layers, may hide the positive transfer. These effects are studied in several network architectures, including VGG16 and ResNet18, on CIFAR10 and ImageNet." @default.
• W3036700096 created "2020-06-25" @default.
• W3036700096 creator A5005429282 @default.
• W3036700096 creator A5014847122 @default.
• W3036700096 creator A5016977101 @default.
• W3036700096 creator A5029133361 @default.
• W3036700096 creator A5041958651 @default.
• W3036700096 creator A5051231311 @default.
• W3036700096 creator A5066294527 @default.
• W3036700096 date "2020-06-18" @default.
• W3036700096 modified "2023-09-27" @default.
• W3036700096 title "What Do Neural Networks Learn When Trained With Random Labels" @default.
• W3036700096 cites W1493372406 @default.
• W3036700096 cites W1514928307 @default.
• W3036700096 cites W1677182931 @default.
• W3036700096 cites W1686810756 @default.
• W3036700096 cites W1836465849 @default.
• W3036700096 cites W1866072925 @default.
• W3036700096 cites W2005067285 @default.
• W3036700096 cites W2017203286 @default.
• W3036700096 cites W2069317438 @default.
• W3036700096 cites W2099579348 @default.
• W3036700096 cites W2108598243 @default.
• W3036700096 cites W2113290770 @default.
• W3036700096 cites W2148349024 @default.
• W3036700096 cites W2149933564 @default.
• W3036700096 cites W2155511299 @default.
• W3036700096 cites W2176412452 @default.
• W3036700096 cites W2302255633 @default.
• W3036700096 cites W2485135680 @default.
• W3036700096 cites W2592335154 @default.
• W3036700096 cites W2593472558 @default.
• W3036700096 cites W2795783309 @default.
• W3036700096 cites W2804773195 @default.
• W3036700096 cites W2889237180 @default.
• W3036700096 cites W2895730082 @default.
• W3036700096 cites W2899385001 @default.
• W3036700096 cites W2911040189 @default.
• W3036700096 cites W2937906318 @default.
• W3036700096 cites W2947556543 @default.
• W3036700096 cites W2962857907 @default.
• W3036700096 cites W2963028646 @default.
• W3036700096 cites W2963081269 @default.
• W3036700096 cites W2963096987 @default.
• W3036700096 cites W2963504252 @default.
• W3036700096 cites W2963664410 @default.
• W3036700096 cites W2963735582 @default.
• W3036700096 cites W2963739978 @default.
• W3036700096 cites W2963759070 @default.
• W3036700096 cites W2963789034 @default.
• W3036700096 cites W2963944430 @default.
• W3036700096 cites W2963959597 @default.
• W3036700096 cites W2964137095 @default.
• W3036700096 cites W2964313743 @default.
• W3036700096 cites W2970854840 @default.
• W3036700096 cites W2970935073 @default.
• W3036700096 cites W2989581875 @default.
• W3036700096 cites W2991040477 @default.
• W3036700096 cites W2996309822 @default.
• W3036700096 cites W3015146382 @default.
• W3036700096 cites W3030163527 @default.
• W3036700096 cites W3118608800 @default.
• W3036700096 cites W3137695714 @default.
• W3036700096 cites W3215037115 @default.
• W3036700096 hasPublicationYear "2020" @default.
• W3036700096 type Work @default.
• W3036700096 sameAs 3036700096 @default.
• W3036700096 citedByCount "12" @default.
• W3036700096 countsByYear W30367000962020 @default.
• W3036700096 countsByYear W30367000962021 @default.
• W3036700096 crossrefType "posted-content" @default.
• W3036700096 hasAuthorship W3036700096A5005429282 @default.
• W3036700096 hasAuthorship W3036700096A5014847122 @default.
• W3036700096 hasAuthorship W3036700096A5016977101 @default.
• W3036700096 hasAuthorship W3036700096A5029133361 @default.
• W3036700096 hasAuthorship W3036700096A5041958651 @default.
• W3036700096 hasAuthorship W3036700096A5051231311 @default.
• W3036700096 hasAuthorship W3036700096A5066294527 @default.
• W3036700096 hasConcept C114614502 @default.
• W3036700096 hasConcept C115961682 @default.
• W3036700096 hasConcept C119857082 @default.
• W3036700096 hasConcept C134306372 @default.
• W3036700096 hasConcept C145420912 @default.
• W3036700096 hasConcept C150899416 @default.
• W3036700096 hasConcept C153180895 @default.
• W3036700096 hasConcept C154945302 @default.
• W3036700096 hasConcept C15744967 @default.
• W3036700096 hasConcept C177148314 @default.
• W3036700096 hasConcept C2524010 @default.
• W3036700096 hasConcept C2780586970 @default.
• W3036700096 hasConcept C30038468 @default.
• W3036700096 hasConcept C33923547 @default.
• W3036700096 hasConcept C41008148 @default.
• W3036700096 hasConcept C45340560 @default.
• W3036700096 hasConcept C50644808 @default.
• W3036700096 hasConcept C77805123 @default.
• W3036700096 hasConcept C81363708 @default.
• W3036700096 hasConcept C99844830 @default.
• W3036700096 hasConceptScore W3036700096C114614502 @default.
• W3036700096 hasConceptScore W3036700096C115961682 @default.

• next