Matches in SemOpenAlex for { <https://semopenalex.org/work/W3213501892> ?p ?o ?g. }
Showing items 1 to 68 of
68
with 100 items per page.
- W3213501892 abstract "Multi-task learning can leverage information learned by one task to benefit the training of other tasks. Despite this capacity, naively training all tasks together in one model often degrades performance, and exhaustively searching through combinations of task groupings can be prohibitively expensive. As a result, efficiently identifying the tasks that would benefit from co-training remains a challenging design question without a clear solution. In this paper, we suggest an approach to select which tasks should train together in multi-task learning models. Our method determines task groupings in a single training run by co-training all tasks together and quantifying the effect to which one task's gradient would affect another task's loss. On the large-scale Taskonomy computer vision dataset, we find this method can decrease test loss by 10.0% compared to simply training all tasks together while operating 11.6 times faster than a state-of-the-art task grouping method." @default.
- W3213501892 created "2021-11-22" @default.
- W3213501892 creator A5005431772 @default.
- W3213501892 creator A5015729655 @default.
- W3213501892 creator A5021315343 @default.
- W3213501892 creator A5038867895 @default.
- W3213501892 creator A5056776503 @default.
- W3213501892 creator A5082659796 @default.
- W3213501892 date "2021-12-06" @default.
- W3213501892 modified "2023-09-25" @default.
- W3213501892 title "Efficiently Identifying Task Groupings for Multi-Task Learning" @default.
- W3213501892 hasPublicationYear "2021" @default.
- W3213501892 type Work @default.
- W3213501892 sameAs 3213501892 @default.
- W3213501892 citedByCount "0" @default.
- W3213501892 crossrefType "proceedings-article" @default.
- W3213501892 hasAuthorship W3213501892A5005431772 @default.
- W3213501892 hasAuthorship W3213501892A5015729655 @default.
- W3213501892 hasAuthorship W3213501892A5021315343 @default.
- W3213501892 hasAuthorship W3213501892A5038867895 @default.
- W3213501892 hasAuthorship W3213501892A5056776503 @default.
- W3213501892 hasAuthorship W3213501892A5082659796 @default.
- W3213501892 hasConcept C119857082 @default.
- W3213501892 hasConcept C127413603 @default.
- W3213501892 hasConcept C153083717 @default.
- W3213501892 hasConcept C154945302 @default.
- W3213501892 hasConcept C175154964 @default.
- W3213501892 hasConcept C201995342 @default.
- W3213501892 hasConcept C2780451532 @default.
- W3213501892 hasConcept C28006648 @default.
- W3213501892 hasConcept C41008148 @default.
- W3213501892 hasConceptScore W3213501892C119857082 @default.
- W3213501892 hasConceptScore W3213501892C127413603 @default.
- W3213501892 hasConceptScore W3213501892C153083717 @default.
- W3213501892 hasConceptScore W3213501892C154945302 @default.
- W3213501892 hasConceptScore W3213501892C175154964 @default.
- W3213501892 hasConceptScore W3213501892C201995342 @default.
- W3213501892 hasConceptScore W3213501892C2780451532 @default.
- W3213501892 hasConceptScore W3213501892C28006648 @default.
- W3213501892 hasConceptScore W3213501892C41008148 @default.
- W3213501892 hasLocation W32135018921 @default.
- W3213501892 hasOpenAccess W3213501892 @default.
- W3213501892 hasPrimaryLocation W32135018921 @default.
- W3213501892 hasRelatedWork W1989663016 @default.
- W3213501892 hasRelatedWork W2056651011 @default.
- W3213501892 hasRelatedWork W2525330092 @default.
- W3213501892 hasRelatedWork W2578423033 @default.
- W3213501892 hasRelatedWork W2753445161 @default.
- W3213501892 hasRelatedWork W2763224350 @default.
- W3213501892 hasRelatedWork W2792402706 @default.
- W3213501892 hasRelatedWork W2895387432 @default.
- W3213501892 hasRelatedWork W2911582157 @default.
- W3213501892 hasRelatedWork W2944900455 @default.
- W3213501892 hasRelatedWork W2965024236 @default.
- W3213501892 hasRelatedWork W2993313557 @default.
- W3213501892 hasRelatedWork W2994972936 @default.
- W3213501892 hasRelatedWork W3094807849 @default.
- W3213501892 hasRelatedWork W3129427614 @default.
- W3213501892 hasRelatedWork W3164719066 @default.
- W3213501892 hasRelatedWork W3176879439 @default.
- W3213501892 hasRelatedWork W3177878514 @default.
- W3213501892 hasRelatedWork W3200030979 @default.
- W3213501892 hasRelatedWork W3208451383 @default.
- W3213501892 hasVolume "34" @default.
- W3213501892 isParatext "false" @default.
- W3213501892 isRetracted "false" @default.
- W3213501892 magId "3213501892" @default.
- W3213501892 workType "article" @default.