Matches in SemOpenAlex for { <https://semopenalex.org/work/W2286376758> ?p ?o ?g. }
Showing items 1 to 93 of
93
with 100 items per page.
- W2286376758 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.
- W2286376758 created "2016-06-24" @default.
- W2286376758 creator A5008894339 @default.
- W2286376758 creator A5020311328 @default.
- W2286376758 creator A5033838234 @default.
- W2286376758 creator A5082808731 @default.
- W2286376758 date "2015-11-20" @default.
- W2286376758 modified "2023-09-27" @default.
- W2286376758 title "Scalable Gradient-Based Tuning of Continuous Regularization Hyperparameters" @default.
- W2286376758 cites W1606347560 @default.
- W2286376758 cites W1864754470 @default.
- W2286376758 cites W1868018859 @default.
- W2286376758 cites W1915968771 @default.
- W2286376758 cites W2006903949 @default.
- W2286376758 cites W2035042171 @default.
- W2286376758 cites W2095705004 @default.
- W2286376758 cites W2097998348 @default.
- W2286376758 cites W2106411961 @default.
- W2286376758 cites W2123045220 @default.
- W2286376758 cites W2130984546 @default.
- W2286376758 cites W2131348505 @default.
- W2286376758 cites W2145094598 @default.
- W2286376758 cites W2152424459 @default.
- W2286376758 cites W2158915909 @default.
- W2286376758 cites W2166107799 @default.
- W2286376758 cites W2335728318 @default.
- W2286376758 cites W2384495648 @default.
- W2286376758 cites W2949117887 @default.
- W2286376758 cites W2964121744 @default.
- W2286376758 cites W3118608800 @default.
- W2286376758 cites W35527955 @default.
- W2286376758 hasPublicationYear "2015" @default.
- W2286376758 type Work @default.
- W2286376758 sameAs 2286376758 @default.
- W2286376758 citedByCount "2" @default.
- W2286376758 countsByYear W22863767582016 @default.
- W2286376758 countsByYear W22863767582019 @default.
- W2286376758 crossrefType "posted-content" @default.
- W2286376758 hasAuthorship W2286376758A5008894339 @default.
- W2286376758 hasAuthorship W2286376758A5020311328 @default.
- W2286376758 hasAuthorship W2286376758A5033838234 @default.
- W2286376758 hasAuthorship W2286376758A5082808731 @default.
- W2286376758 hasConcept C10485038 @default.
- W2286376758 hasConcept C119857082 @default.
- W2286376758 hasConcept C12267149 @default.
- W2286376758 hasConcept C154945302 @default.
- W2286376758 hasConcept C190502265 @default.
- W2286376758 hasConcept C2776135515 @default.
- W2286376758 hasConcept C41008148 @default.
- W2286376758 hasConcept C48044578 @default.
- W2286376758 hasConcept C50644808 @default.
- W2286376758 hasConcept C77088390 @default.
- W2286376758 hasConcept C8642999 @default.
- W2286376758 hasConcept C93959086 @default.
- W2286376758 hasConceptScore W2286376758C10485038 @default.
- W2286376758 hasConceptScore W2286376758C119857082 @default.
- W2286376758 hasConceptScore W2286376758C12267149 @default.
- W2286376758 hasConceptScore W2286376758C154945302 @default.
- W2286376758 hasConceptScore W2286376758C190502265 @default.
- W2286376758 hasConceptScore W2286376758C2776135515 @default.
- W2286376758 hasConceptScore W2286376758C41008148 @default.
- W2286376758 hasConceptScore W2286376758C48044578 @default.
- W2286376758 hasConceptScore W2286376758C50644808 @default.
- W2286376758 hasConceptScore W2286376758C77088390 @default.
- W2286376758 hasConceptScore W2286376758C8642999 @default.
- W2286376758 hasConceptScore W2286376758C93959086 @default.
- W2286376758 hasLocation W22863767581 @default.
- W2286376758 hasOpenAccess W2286376758 @default.
- W2286376758 hasPrimaryLocation W22863767581 @default.
- W2286376758 hasRelatedWork W1519208986 @default.
- W2286376758 hasRelatedWork W1582527637 @default.
- W2286376758 hasRelatedWork W1998039793 @default.
- W2286376758 hasRelatedWork W2016073575 @default.
- W2286376758 hasRelatedWork W2120060023 @default.
- W2286376758 hasRelatedWork W2128548761 @default.
- W2286376758 hasRelatedWork W2176767885 @default.
- W2286376758 hasRelatedWork W2397493451 @default.
- W2286376758 hasRelatedWork W2735749309 @default.
- W2286376758 hasRelatedWork W2937590067 @default.
- W2286376758 hasRelatedWork W2941876110 @default.
- W2286376758 hasRelatedWork W2963565814 @default.
- W2286376758 hasRelatedWork W2964029277 @default.
- W2286376758 hasRelatedWork W2984907890 @default.
- W2286376758 hasRelatedWork W3015579318 @default.
- W2286376758 hasRelatedWork W3081201943 @default.
- W2286376758 hasRelatedWork W3116576507 @default.
- W2286376758 hasRelatedWork W3118861876 @default.
- W2286376758 hasRelatedWork W3134608380 @default.
- W2286376758 hasRelatedWork W3175497641 @default.
- W2286376758 isParatext "false" @default.
- W2286376758 isRetracted "false" @default.
- W2286376758 magId "2286376758" @default.
- W2286376758 workType "article" @default.