Matches in SemOpenAlex for { <https://semopenalex.org/work/W1760458529> ?p ?o ?g. }
Showing items 1 to 96 of
96
with 100 items per page.
- W1760458529 abstract "Training deep belief networks (DBNs) requires optimizing a non-convex function with an extremely large number of parameters. Naturally, existing gradient descent (GD) based methods are prone to arbitrarily poor local minima. In this paper, we rigorously show that such local minima can be avoided (upto an approximation error) by using the dropout technique, a widely used heuristic in this domain. In particular, we show that by randomly dropping a few nodes of a one-hidden layer neural network, the training objective function, up to a certain approximation error, decreases by a multiplicative factor. On the flip side, we show that for training convex empirical risk minimizers (ERM), dropout in fact acts as a stabilizer or regularizer. That is, a simple dropout based GD method for convex ERMs is stable in the face of arbitrary changes to any one of the training points. Using the above assertion, we show that dropout provides fast rates for generalization error in learning (convex) generalized linear models (GLM). Moreover, using the above mentioned stability properties of dropout, we design dropout based differentially private algorithms for solving ERMs. The learned GLM thus, preserves privacy of each of the individual training points while providing accurate predictions for new test points. Finally, we empirically validate our stability assertions for dropout in the context of convex ERMs and show that surprisingly, dropout significantly outperforms (in terms of prediction accuracy) the L2 regularization based methods for several benchmark datasets." @default.
- W1760458529 created "2016-06-24" @default.
- W1760458529 creator A5005434490 @default.
- W1760458529 creator A5008213441 @default.
- W1760458529 creator A5013097792 @default.
- W1760458529 creator A5034432097 @default.
- W1760458529 date "2015-03-06" @default.
- W1760458529 modified "2023-09-24" @default.
- W1760458529 title "To Drop or Not to Drop: Robustness, Consistency and Differential Privacy Properties of Dropout" @default.
- W1760458529 cites W119676333 @default.
- W1760458529 cites W1516903196 @default.
- W1760458529 cites W1622633983 @default.
- W1760458529 cites W1673923490 @default.
- W1760458529 cites W1746889935 @default.
- W1760458529 cites W1873763122 @default.
- W1760458529 cites W1904365287 @default.
- W1760458529 cites W1985511977 @default.
- W1760458529 cites W2053801139 @default.
- W1760458529 cites W2064868049 @default.
- W1760458529 cites W2109426455 @default.
- W1760458529 cites W2112380340 @default.
- W1760458529 cites W2113517874 @default.
- W1760458529 cites W2114609025 @default.
- W1760458529 cites W2119874464 @default.
- W1760458529 cites W2135754715 @default.
- W1760458529 cites W2136351519 @default.
- W1760458529 cites W2138865266 @default.
- W1760458529 cites W2139338362 @default.
- W1760458529 cites W2158542502 @default.
- W1760458529 cites W2164370980 @default.
- W1760458529 cites W2250968750 @default.
- W1760458529 cites W2911978475 @default.
- W1760458529 cites W2949176378 @default.
- W1760458529 cites W2952825952 @default.
- W1760458529 cites W3120740533 @default.
- W1760458529 cites W35527955 @default.
- W1760458529 hasPublicationYear "2015" @default.
- W1760458529 type Work @default.
- W1760458529 sameAs 1760458529 @default.
- W1760458529 citedByCount "11" @default.
- W1760458529 countsByYear W17604585292016 @default.
- W1760458529 countsByYear W17604585292017 @default.
- W1760458529 countsByYear W17604585292019 @default.
- W1760458529 countsByYear W17604585292020 @default.
- W1760458529 crossrefType "posted-content" @default.
- W1760458529 hasAuthorship W1760458529A5005434490 @default.
- W1760458529 hasAuthorship W1760458529A5008213441 @default.
- W1760458529 hasAuthorship W1760458529A5013097792 @default.
- W1760458529 hasAuthorship W1760458529A5034432097 @default.
- W1760458529 hasConcept C107321475 @default.
- W1760458529 hasConcept C119857082 @default.
- W1760458529 hasConcept C126255220 @default.
- W1760458529 hasConcept C134306372 @default.
- W1760458529 hasConcept C154945302 @default.
- W1760458529 hasConcept C186633575 @default.
- W1760458529 hasConcept C2776145597 @default.
- W1760458529 hasConcept C28826006 @default.
- W1760458529 hasConcept C33923547 @default.
- W1760458529 hasConcept C41008148 @default.
- W1760458529 hasConceptScore W1760458529C107321475 @default.
- W1760458529 hasConceptScore W1760458529C119857082 @default.
- W1760458529 hasConceptScore W1760458529C126255220 @default.
- W1760458529 hasConceptScore W1760458529C134306372 @default.
- W1760458529 hasConceptScore W1760458529C154945302 @default.
- W1760458529 hasConceptScore W1760458529C186633575 @default.
- W1760458529 hasConceptScore W1760458529C2776145597 @default.
- W1760458529 hasConceptScore W1760458529C28826006 @default.
- W1760458529 hasConceptScore W1760458529C33923547 @default.
- W1760458529 hasConceptScore W1760458529C41008148 @default.
- W1760458529 hasLocation W17604585291 @default.
- W1760458529 hasOpenAccess W1760458529 @default.
- W1760458529 hasPrimaryLocation W17604585291 @default.
- W1760458529 hasRelatedWork W1746889935 @default.
- W1760458529 hasRelatedWork W2027595342 @default.
- W1760458529 hasRelatedWork W2095705004 @default.
- W1760458529 hasRelatedWork W2110091014 @default.
- W1760458529 hasRelatedWork W2112796928 @default.
- W1760458529 hasRelatedWork W2182267394 @default.
- W1760458529 hasRelatedWork W2473418344 @default.
- W1760458529 hasRelatedWork W2523888555 @default.
- W1760458529 hasRelatedWork W2535690855 @default.
- W1760458529 hasRelatedWork W2786968269 @default.
- W1760458529 hasRelatedWork W2944145193 @default.
- W1760458529 hasRelatedWork W2952550984 @default.
- W1760458529 hasRelatedWork W2952825952 @default.
- W1760458529 hasRelatedWork W2963918617 @default.
- W1760458529 hasRelatedWork W3034384909 @default.
- W1760458529 hasRelatedWork W3084926902 @default.
- W1760458529 hasRelatedWork W3099141078 @default.
- W1760458529 hasRelatedWork W3110712518 @default.
- W1760458529 hasRelatedWork W3118608800 @default.
- W1760458529 hasRelatedWork W35527955 @default.
- W1760458529 isParatext "false" @default.
- W1760458529 isRetracted "false" @default.
- W1760458529 magId "1760458529" @default.
- W1760458529 workType "article" @default.