FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2950673314> ?p ?o ?g. }

• W2950673314 abstract "In multi-task learning, multiple tasks are solved jointly, sharing inductive bias between them. Multi-task learning is inherently a multi-objective problem because different tasks may conflict, necessitating a trade-off. A common compromise is to optimize a proxy objective that minimizes a weighted linear combination of per-task losses. However, this workaround is only valid when the tasks do not compete, which is rarely the case. In this paper, we explicitly cast multi-task learning as multi-objective optimization, with the overall objective of finding a Pareto optimal solution. To this end, we use algorithms developed in the gradient-based multi-objective optimization literature. These algorithms are not directly applicable to large-scale learning problems since they scale poorly with the dimensionality of the gradients and the number of tasks. We therefore propose an upper bound for the multi-objective loss and show that it can be optimized efficiently. We further prove that optimizing this upper bound yields a Pareto optimal solution under realistic assumptions. We apply our method to a variety of multi-task deep learning problems including digit classification, scene understanding (joint semantic segmentation, instance segmentation, and depth estimation), and multi-label classification. Our method produces higher-performing models than recent multi-task learning formulations or per-task training." @default.
• W2950673314 created "2019-06-27" @default.
• W2950673314 creator A5006181255 @default.
• W2950673314 creator A5026399398 @default.
• W2950673314 date "2018-10-10" @default.
• W2950673314 modified "2023-09-23" @default.
• W2950673314 title "Multi-Task Learning as Multi-Objective Optimization" @default.
• W2950673314 cites W1572469720 @default.
• W2950673314 cites W1598266570 @default.
• W2950673314 cites W1834627138 @default.
• W2950673314 cites W1988210060 @default.
• W2950673314 cites W1992586477 @default.
• W2950673314 cites W2008499862 @default.
• W2950673314 cites W2025198378 @default.
• W2950673314 cites W2025747217 @default.
• W2950673314 cites W2060727197 @default.
• W2950673314 cites W2085291930 @default.
• W2950673314 cites W2094035326 @default.
• W2950673314 cites W2094911742 @default.
• W2950673314 cites W2100717205 @default.
• W2950673314 cites W2108598243 @default.
• W2950673314 cites W2112796928 @default.
• W2950673314 cites W2117130368 @default.
• W2950673314 cites W2118099552 @default.
• W2950673314 cites W2119187866 @default.
• W2950673314 cites W2129759879 @default.
• W2950673314 cites W2131479143 @default.
• W2950673314 cites W2141084846 @default.
• W2950673314 cites W2159524357 @default.
• W2950673314 cites W2165644552 @default.
• W2950673314 cites W2172589779 @default.
• W2950673314 cites W2194775991 @default.
• W2950673314 cites W2221104225 @default.
• W2950673314 cites W2251743902 @default.
• W2950673314 cites W2295072214 @default.
• W2950673314 cites W2339391301 @default.
• W2950673314 cites W2407793339 @default.
• W2950673314 cites W2432541215 @default.
• W2950673314 cites W2467110937 @default.
• W2950673314 cites W2556468274 @default.
• W2950673314 cites W2560023338 @default.
• W2950673314 cites W2562989799 @default.
• W2950673314 cites W2615258659 @default.
• W2950673314 cites W2624871570 @default.
• W2950673314 cites W2626792426 @default.
• W2950673314 cites W2727285023 @default.
• W2950673314 cites W2737258237 @default.
• W2950673314 cites W2767175863 @default.
• W2950673314 cites W2769783658 @default.
• W2950673314 cites W2798512429 @default.
• W2950673314 cites W2798766386 @default.
• W2950673314 cites W2899771611 @default.
• W2950673314 cites W2913340405 @default.
• W2950673314 cites W2953139137 @default.
• W2950673314 cites W2962743139 @default.
• W2950673314 cites W2963498646 @default.
• W2950673314 cites W2963677766 @default.
• W2950673314 cites W2963703618 @default.
• W2950673314 cites W2964056935 @default.
• W2950673314 cites W2964121793 @default.
• W2950673314 cites W582055897 @default.
• W2950673314 cites W905619 @default.
• W2950673314 hasPublicationYear "2018" @default.
• W2950673314 type Work @default.
• W2950673314 sameAs 2950673314 @default.
• W2950673314 citedByCount "30" @default.
• W2950673314 countsByYear W29506733142017 @default.
• W2950673314 countsByYear W29506733142018 @default.
• W2950673314 countsByYear W29506733142019 @default.
• W2950673314 countsByYear W29506733142020 @default.
• W2950673314 countsByYear W29506733142021 @default.
• W2950673314 crossrefType "posted-content" @default.
• W2950673314 hasAuthorship W2950673314A5006181255 @default.
• W2950673314 hasAuthorship W2950673314A5026399398 @default.
• W2950673314 hasConcept C111030470 @default.
• W2950673314 hasConcept C11413529 @default.
• W2950673314 hasConcept C119857082 @default.
• W2950673314 hasConcept C126255220 @default.
• W2950673314 hasConcept C137635306 @default.
• W2950673314 hasConcept C137836250 @default.
• W2950673314 hasConcept C154945302 @default.
• W2950673314 hasConcept C162324750 @default.
• W2950673314 hasConcept C187736073 @default.
• W2950673314 hasConcept C194541083 @default.
• W2950673314 hasConcept C197352929 @default.
• W2950673314 hasConcept C199360897 @default.
• W2950673314 hasConcept C2780451532 @default.
• W2950673314 hasConcept C28006648 @default.
• W2950673314 hasConcept C33923547 @default.
• W2950673314 hasConcept C41008148 @default.
• W2950673314 hasConcept C89600930 @default.
• W2950673314 hasConceptScore W2950673314C111030470 @default.
• W2950673314 hasConceptScore W2950673314C11413529 @default.
• W2950673314 hasConceptScore W2950673314C119857082 @default.
• W2950673314 hasConceptScore W2950673314C126255220 @default.
• W2950673314 hasConceptScore W2950673314C137635306 @default.
• W2950673314 hasConceptScore W2950673314C137836250 @default.
• W2950673314 hasConceptScore W2950673314C154945302 @default.
• W2950673314 hasConceptScore W2950673314C162324750 @default.
• W2950673314 hasConceptScore W2950673314C187736073 @default.

• next