FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2982206528> ?p ?o ?g. }

• W2982206528 abstract "We propose in this paper a general framework for deriving loss functions for structured prediction. In our framework, the user chooses a convex set including the output space and provides an oracle for projecting onto that set. Given that oracle, our framework automatically generates a corresponding convex and smooth loss function. As we show, adding a projection as output layer provably makes the loss smaller. We identify the marginal polytope, the output space's convex hull, as the best convex set on which to project. However, because the projection onto the marginal polytope can sometimes be expensive to compute, we allow to use any convex superset instead, with potentially cheaper-to-compute projection. Since efficient projection algorithms are available for numerous convex sets, this allows us to construct loss functions for a variety of tasks. On the theoretical side, when combined with calibrated decoding, we prove that our loss functions can be used as a consistent surrogate for a (potentially non-convex) target loss function of interest. We demonstrate our losses on label ranking, ordinal regression and multilabel classification, confirming the improved accuracy enabled by projections." @default.
• W2982206528 created "2019-11-01" @default.
• W2982206528 creator A5049123454 @default.
• W2982206528 date "2019-10-24" @default.
• W2982206528 modified "2023-09-27" @default.
• W2982206528 title "Structured Prediction with Projection Oracles" @default.
• W2982206528 cites W1480538416 @default.
• W2982206528 cites W1511358834 @default.
• W2982206528 cites W1515020792 @default.
• W2982206528 cites W159752407 @default.
• W2982206528 cites W1982032418 @default.
• W2982206528 cites W1983905049 @default.
• W2982206528 cites W2003447360 @default.
• W2982206528 cites W2006912660 @default.
• W2982206528 cites W2008652694 @default.
• W2982206528 cites W2023163512 @default.
• W2982206528 cites W2061211758 @default.
• W2982206528 cites W2093781566 @default.
• W2982206528 cites W2096765209 @default.
• W2982206528 cites W2097826433 @default.
• W2982206528 cites W2101234009 @default.
• W2982206528 cites W2103194807 @default.
• W2982206528 cites W2105842272 @default.
• W2982206528 cites W2114028889 @default.
• W2982206528 cites W2120340025 @default.
• W2982206528 cites W2140472169 @default.
• W2982206528 cites W2141407058 @default.
• W2982206528 cites W2147880316 @default.
• W2982206528 cites W2155344786 @default.
• W2982206528 cites W2155434396 @default.
• W2982206528 cites W2186629860 @default.
• W2982206528 cites W2187061624 @default.
• W2982206528 cites W2188936328 @default.
• W2982206528 cites W2190044943 @default.
• W2982206528 cites W2222512263 @default.
• W2982206528 cites W2252239149 @default.
• W2982206528 cites W2295830385 @default.
• W2982206528 cites W2296319761 @default.
• W2982206528 cites W2403513640 @default.
• W2982206528 cites W2409290336 @default.
• W2982206528 cites W2477742853 @default.
• W2982206528 cites W2538817717 @default.
• W2982206528 cites W2593397662 @default.
• W2982206528 cites W2758087204 @default.
• W2982206528 cites W2876724543 @default.
• W2982206528 cites W2897802964 @default.
• W2982206528 cites W2909691537 @default.
• W2982206528 cites W2914644247 @default.
• W2982206528 cites W2918249320 @default.
• W2982206528 cites W2946472767 @default.
• W2982206528 cites W2950304314 @default.
• W2982206528 cites W2950492466 @default.
• W2982206528 cites W2962736834 @default.
• W2982206528 cites W2962839844 @default.
• W2982206528 cites W2962902266 @default.
• W2982206528 cites W2963123301 @default.
• W2982206528 cites W2963343251 @default.
• W2982206528 cites W2964076774 @default.
• W2982206528 cites W2964095938 @default.
• W2982206528 cites W2964168821 @default.
• W2982206528 cites W2964231757 @default.
• W2982206528 cites W30074706 @default.
• W2982206528 cites W3022720367 @default.
• W2982206528 hasPublicationYear "2019" @default.
• W2982206528 type Work @default.
• W2982206528 sameAs 2982206528 @default.
• W2982206528 citedByCount "0" @default.
• W2982206528 crossrefType "posted-content" @default.
• W2982206528 hasAuthorship W2982206528A5049123454 @default.
• W2982206528 hasConcept C110943637 @default.
• W2982206528 hasConcept C111110010 @default.
• W2982206528 hasConcept C112680207 @default.
• W2982206528 hasConcept C11413529 @default.
• W2982206528 hasConcept C114614502 @default.
• W2982206528 hasConcept C115903868 @default.
• W2982206528 hasConcept C12108790 @default.
• W2982206528 hasConcept C126255220 @default.
• W2982206528 hasConcept C14036430 @default.
• W2982206528 hasConcept C145691206 @default.
• W2982206528 hasConcept C154945302 @default.
• W2982206528 hasConcept C157972887 @default.
• W2982206528 hasConcept C177264268 @default.
• W2982206528 hasConcept C189430467 @default.
• W2982206528 hasConcept C199360897 @default.
• W2982206528 hasConcept C206194317 @default.
• W2982206528 hasConcept C2524010 @default.
• W2982206528 hasConcept C33923547 @default.
• W2982206528 hasConcept C41008148 @default.
• W2982206528 hasConcept C49870271 @default.
• W2982206528 hasConcept C50862404 @default.
• W2982206528 hasConcept C55166926 @default.
• W2982206528 hasConcept C57493831 @default.
• W2982206528 hasConcept C78458016 @default.
• W2982206528 hasConcept C86803240 @default.
• W2982206528 hasConceptScore W2982206528C110943637 @default.
• W2982206528 hasConceptScore W2982206528C111110010 @default.
• W2982206528 hasConceptScore W2982206528C112680207 @default.
• W2982206528 hasConceptScore W2982206528C11413529 @default.
• W2982206528 hasConceptScore W2982206528C114614502 @default.
• W2982206528 hasConceptScore W2982206528C115903868 @default.

• next