FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2964029277> ?p ?o ?g. }

• W2964029277 endingPage "2960" @default.
• W2964029277 startingPage "2952" @default.
• W2964029277 abstract "Hyperparameter selection generally relies on running multiple full training trials, with selection based on validation set performance. We propose a gradient-based approach for locally adjusting hyperparameters during training of the model. Hyperparameters are adjusted so as to make the model parameter gradients, and hence updates, more advantageous for the validation cost. We explore the approach for tuning regularization hyperparameters and find that in experiments on MNIST, SVHN and CIFAR-10, the resulting regularization levels are within the optimal regions. The additional computational cost depends on how frequently the hyperparameters are trained, but the tested scheme adds only 30% computational overhead regardless of the model size. Since the method is significantly less computationally demanding compared to similar gradient-based approaches to hyperparameter optimization, and consistently finds good hyperparameter values, it can be a useful tool for training neural network models." @default.
• W2964029277 created "2019-07-30" @default.
• W2964029277 creator A5008894339 @default.
• W2964029277 creator A5020311328 @default.
• W2964029277 creator A5033838234 @default.
• W2964029277 creator A5082808731 @default.
• W2964029277 date "2016-06-19" @default.
• W2964029277 modified "2023-09-23" @default.
• W2964029277 title "Scalable gradient-based tuning of continuous regularization hyperparameters" @default.
• W2964029277 cites W1864754470 @default.
• W2964029277 cites W1868018859 @default.
• W2964029277 cites W1915968771 @default.
• W2964029277 cites W1921523184 @default.
• W2964029277 cites W2006903949 @default.
• W2964029277 cites W2035042171 @default.
• W2964029277 cites W2062227835 @default.
• W2964029277 cites W2095705004 @default.
• W2964029277 cites W2097998348 @default.
• W2964029277 cites W2106411961 @default.
• W2964029277 cites W2123045220 @default.
• W2964029277 cites W2130984546 @default.
• W2964029277 cites W2131348505 @default.
• W2964029277 cites W2145094598 @default.
• W2964029277 cites W2152424459 @default.
• W2964029277 cites W2158915909 @default.
• W2964029277 cites W2166107799 @default.
• W2964029277 cites W2335728318 @default.
• W2964029277 cites W2384495648 @default.
• W2964029277 cites W2949117887 @default.
• W2964029277 cites W2964121744 @default.
• W2964029277 cites W3118608800 @default.
• W2964029277 cites W35527955 @default.
• W2964029277 cites W830076066 @default.
• W2964029277 hasPublicationYear "2016" @default.
• W2964029277 type Work @default.
• W2964029277 sameAs 2964029277 @default.
• W2964029277 citedByCount "55" @default.
• W2964029277 countsByYear W29640292772016 @default.
• W2964029277 countsByYear W29640292772018 @default.
• W2964029277 countsByYear W29640292772019 @default.
• W2964029277 countsByYear W29640292772020 @default.
• W2964029277 countsByYear W29640292772021 @default.
• W2964029277 countsByYear W29640292772022 @default.
• W2964029277 crossrefType "proceedings-article" @default.
• W2964029277 hasAuthorship W2964029277A5008894339 @default.
• W2964029277 hasAuthorship W2964029277A5020311328 @default.
• W2964029277 hasAuthorship W2964029277A5033838234 @default.
• W2964029277 hasAuthorship W2964029277A5082808731 @default.
• W2964029277 hasConcept C10485038 @default.
• W2964029277 hasConcept C119857082 @default.
• W2964029277 hasConcept C12267149 @default.
• W2964029277 hasConcept C154945302 @default.
• W2964029277 hasConcept C190502265 @default.
• W2964029277 hasConcept C2776135515 @default.
• W2964029277 hasConcept C41008148 @default.
• W2964029277 hasConcept C48044578 @default.
• W2964029277 hasConcept C50644808 @default.
• W2964029277 hasConcept C77088390 @default.
• W2964029277 hasConcept C8642999 @default.
• W2964029277 hasConcept C93959086 @default.
• W2964029277 hasConceptScore W2964029277C10485038 @default.
• W2964029277 hasConceptScore W2964029277C119857082 @default.
• W2964029277 hasConceptScore W2964029277C12267149 @default.
• W2964029277 hasConceptScore W2964029277C154945302 @default.
• W2964029277 hasConceptScore W2964029277C190502265 @default.
• W2964029277 hasConceptScore W2964029277C2776135515 @default.
• W2964029277 hasConceptScore W2964029277C41008148 @default.
• W2964029277 hasConceptScore W2964029277C48044578 @default.
• W2964029277 hasConceptScore W2964029277C50644808 @default.
• W2964029277 hasConceptScore W2964029277C77088390 @default.
• W2964029277 hasConceptScore W2964029277C8642999 @default.
• W2964029277 hasConceptScore W2964029277C93959086 @default.
• W2964029277 hasLocation W29640292771 @default.
• W2964029277 hasOpenAccess W2964029277 @default.
• W2964029277 hasPrimaryLocation W29640292771 @default.
• W2964029277 hasRelatedWork W2097998348 @default.
• W2964029277 hasRelatedWork W2106411961 @default.
• W2964029277 hasRelatedWork W2131241448 @default.
• W2964029277 hasRelatedWork W2166107799 @default.
• W2964029277 hasRelatedWork W2187061624 @default.
• W2964029277 hasRelatedWork W2194775991 @default.
• W2964029277 hasRelatedWork W2286376758 @default.
• W2964029277 hasRelatedWork W2604763608 @default.
• W2964029277 hasRelatedWork W2753160622 @default.
• W2964029277 hasRelatedWork W2786022758 @default.
• W2964029277 hasRelatedWork W2810075754 @default.
• W2964029277 hasRelatedWork W2962826047 @default.
• W2964029277 hasRelatedWork W2963233958 @default.
• W2964029277 hasRelatedWork W2963306862 @default.
• W2964029277 hasRelatedWork W2963804140 @default.
• W2964029277 hasRelatedWork W2963815651 @default.
• W2964029277 hasRelatedWork W2964121744 @default.
• W2964029277 hasRelatedWork W3037120332 @default.
• W2964029277 hasRelatedWork W3118608800 @default.
• W2964029277 hasRelatedWork W60686164 @default.
• W2964029277 isParatext "false" @default.
• W2964029277 isRetracted "false" @default.
• W2964029277 magId "2964029277" @default.

• next