FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3110320718> ?p ?o ?g. }

• W3110320718 abstract "Many problems require to optimize empirical risk functions over large data sets. Gradient descent methods that calculate the full gradient in every descent step do not scale to such datasets. Various flavours of Stochastic Gradient Descent (SGD) replace the expensive summation that computes the full gradient by approximating it with a small sum over a randomly selected subsample of the data set that in turn suffers from a high variance. We present a different approach that is inspired by classical results of Tchakaloff and Caratheodory about measure reduction. These results allow to replace an empirical measure with another, carefully constructed probability measure that has a much smaller support, but can preserve certain statistics such as the expected gradient. To turn this into scalable algorithms we firstly, adaptively select the descent steps where the measure reduction is carried out; secondly, we combine this with Block Coordinate Descent so that measure reduction can be done very cheaply. This makes the resulting methods scalable to high-dimensional spaces. Finally, we provide an experimental validation and comparison." @default.
• W3110320718 created "2020-12-07" @default.
• W3110320718 creator A5024452289 @default.
• W3110320718 creator A5077632031 @default.
• W3110320718 creator A5091718585 @default.
• W3110320718 date "2020-06-02" @default.
• W3110320718 modified "2023-09-27" @default.
• W3110320718 title "Carath'eodory Sampling for Stochastic Gradient Descent" @default.
• W3110320718 cites W114517082 @default.
• W3110320718 cites W1503398984 @default.
• W3110320718 cites W1522301498 @default.
• W3110320718 cites W1947202642 @default.
• W3110320718 cites W1968154520 @default.
• W3110320718 cites W1988795359 @default.
• W3110320718 cites W1994616650 @default.
• W3110320718 cites W2000769684 @default.
• W3110320718 cites W2020098476 @default.
• W3110320718 cites W2023901033 @default.
• W3110320718 cites W2032395696 @default.
• W3110320718 cites W2039050532 @default.
• W3110320718 cites W2052963818 @default.
• W3110320718 cites W2072265605 @default.
• W3110320718 cites W2079556846 @default.
• W3110320718 cites W2086161653 @default.
• W3110320718 cites W2095984592 @default.
• W3110320718 cites W2096725512 @default.
• W3110320718 cites W2102486516 @default.
• W3110320718 cites W2105875671 @default.
• W3110320718 cites W2106398669 @default.
• W3110320718 cites W2114643899 @default.
• W3110320718 cites W2117229703 @default.
• W3110320718 cites W2117575248 @default.
• W3110320718 cites W2135046866 @default.
• W3110320718 cites W2144070955 @default.
• W3110320718 cites W2164075197 @default.
• W3110320718 cites W2164641162 @default.
• W3110320718 cites W23064316 @default.
• W3110320718 cites W2523246573 @default.
• W3110320718 cites W2553080076 @default.
• W3110320718 cites W2556206132 @default.
• W3110320718 cites W2616811598 @default.
• W3110320718 cites W2753969545 @default.
• W3110320718 cites W2760299880 @default.
• W3110320718 cites W2777891128 @default.
• W3110320718 cites W2949557549 @default.
• W3110320718 cites W2951028711 @default.
• W3110320718 cites W2951801506 @default.
• W3110320718 cites W2962727278 @default.
• W3110320718 cites W2962990180 @default.
• W3110320718 cites W2963433607 @default.
• W3110320718 cites W2963434703 @default.
• W3110320718 cites W2963707635 @default.
• W3110320718 cites W2964013670 @default.
• W3110320718 cites W2964106020 @default.
• W3110320718 cites W3099937901 @default.
• W3110320718 cites W3101306630 @default.
• W3110320718 cites W3101608030 @default.
• W3110320718 cites W816981004 @default.
• W3110320718 cites W2963821730 @default.
• W3110320718 hasPublicationYear "2020" @default.
• W3110320718 type Work @default.
• W3110320718 sameAs 3110320718 @default.
• W3110320718 citedByCount "1" @default.
• W3110320718 countsByYear W31103207182020 @default.
• W3110320718 crossrefType "posted-content" @default.
• W3110320718 hasAuthorship W3110320718A5024452289 @default.
• W3110320718 hasAuthorship W3110320718A5077632031 @default.
• W3110320718 hasAuthorship W3110320718A5091718585 @default.
• W3110320718 hasConcept C105795698 @default.
• W3110320718 hasConcept C106131492 @default.
• W3110320718 hasConcept C111335779 @default.
• W3110320718 hasConcept C11413529 @default.
• W3110320718 hasConcept C121955636 @default.
• W3110320718 hasConcept C124101348 @default.
• W3110320718 hasConcept C126255220 @default.
• W3110320718 hasConcept C127413603 @default.
• W3110320718 hasConcept C140779682 @default.
• W3110320718 hasConcept C144133560 @default.
• W3110320718 hasConcept C146978453 @default.
• W3110320718 hasConcept C153258448 @default.
• W3110320718 hasConcept C154945302 @default.
• W3110320718 hasConcept C157553263 @default.
• W3110320718 hasConcept C177264268 @default.
• W3110320718 hasConcept C19499675 @default.
• W3110320718 hasConcept C196083921 @default.
• W3110320718 hasConcept C199360897 @default.
• W3110320718 hasConcept C206688291 @default.
• W3110320718 hasConcept C2524010 @default.
• W3110320718 hasConcept C2776637919 @default.
• W3110320718 hasConcept C2780009758 @default.
• W3110320718 hasConcept C31972630 @default.
• W3110320718 hasConcept C33923547 @default.
• W3110320718 hasConcept C41008148 @default.
• W3110320718 hasConcept C48044578 @default.
• W3110320718 hasConcept C50644808 @default.
• W3110320718 hasConcept C62644790 @default.
• W3110320718 hasConcept C77088390 @default.
• W3110320718 hasConceptScore W3110320718C105795698 @default.
• W3110320718 hasConceptScore W3110320718C106131492 @default.
• W3110320718 hasConceptScore W3110320718C111335779 @default.

• next