Matches in SemOpenAlex for { <https://semopenalex.org/work/W2990147108> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W2990147108 abstract "Deep learning models are often trained on datasets that contain sensitive information such as individuals' shopping transactions, personal contacts, and medical records. An increasingly important line of work therefore has sought to train neural networks subject to privacy constraints that are specified by differential privacy or its divergence-based relaxations. These privacy definitions, however, have weaknesses in handling certain important primitives (composition and subsampling), thereby giving loose or complicated privacy analyses of training neural networks. In this paper, we consider a recently proposed privacy definition termed textit{$f$-differential privacy} [18] for a refined privacy analysis of training neural networks. Leveraging the appealing properties of $f$-differential privacy in handling composition and subsampling, this paper derives analytically tractable expressions for the privacy guarantees of both stochastic gradient descent and Adam used in training deep neural networks, without the need of developing sophisticated techniques as [3] did. Our results demonstrate that the $f$-differential privacy framework allows for a new privacy analysis that improves on the prior analysis~[3], which in turn suggests tuning certain parameters of neural networks for a better prediction accuracy without violating the privacy budget. These theoretically derived improvements are confirmed by our experiments in a range of tasks in image classification, text classification, and recommender systems. Python code to calculate the privacy cost for these experiments is publicly available in the texttt{TensorFlow Privacy} library." @default.
- W2990147108 created "2019-12-05" @default.
- W2990147108 creator A5002149616 @default.
- W2990147108 creator A5020028434 @default.
- W2990147108 creator A5076817901 @default.
- W2990147108 creator A5080575294 @default.
- W2990147108 date "2019-11-26" @default.
- W2990147108 modified "2023-09-23" @default.
- W2990147108 title "Deep Learning with Gaussian Differential Privacy" @default.
- W2990147108 cites W1504694836 @default.
- W2990147108 cites W1507034068 @default.
- W2990147108 cites W1522301498 @default.
- W2990147108 cites W1557833142 @default.
- W2990147108 cites W1873763122 @default.
- W2990147108 cites W1992926795 @default.
- W2990147108 cites W1997690112 @default.
- W2990147108 cites W2053637704 @default.
- W2990147108 cites W2064675550 @default.
- W2990147108 cites W2113459411 @default.
- W2990147108 cites W2122111042 @default.
- W2990147108 cites W2135930857 @default.
- W2990147108 cites W2146502635 @default.
- W2990147108 cites W2149706766 @default.
- W2990147108 cites W2153635508 @default.
- W2990147108 cites W2219888463 @default.
- W2990147108 cites W2294904676 @default.
- W2990147108 cites W2605258322 @default.
- W2990147108 cites W2753855453 @default.
- W2990147108 cites W2785361959 @default.
- W2990147108 cites W2809008139 @default.
- W2990147108 cites W2905148628 @default.
- W2990147108 cites W2943912735 @default.
- W2990147108 cites W2946252494 @default.
- W2990147108 cites W2950527268 @default.
- W2990147108 cites W2950602864 @default.
- W2990147108 cites W2963543276 @default.
- W2990147108 cites W2976201062 @default.
- W2990147108 cites W3102407811 @default.
- W2990147108 cites W3120740533 @default.
- W2990147108 doi "https://doi.org/10.48550/arxiv.1911.11607" @default.
- W2990147108 hasPublicationYear "2019" @default.
- W2990147108 type Work @default.
- W2990147108 sameAs 2990147108 @default.
- W2990147108 citedByCount "13" @default.
- W2990147108 countsByYear W29901471082019 @default.
- W2990147108 countsByYear W29901471082020 @default.
- W2990147108 countsByYear W29901471082021 @default.
- W2990147108 crossrefType "posted-content" @default.
- W2990147108 hasAuthorship W2990147108A5002149616 @default.
- W2990147108 hasAuthorship W2990147108A5020028434 @default.
- W2990147108 hasAuthorship W2990147108A5076817901 @default.
- W2990147108 hasAuthorship W2990147108A5080575294 @default.
- W2990147108 hasBestOaLocation W29901471081 @default.
- W2990147108 hasConcept C108583219 @default.
- W2990147108 hasConcept C119857082 @default.
- W2990147108 hasConcept C124101348 @default.
- W2990147108 hasConcept C154945302 @default.
- W2990147108 hasConcept C23130292 @default.
- W2990147108 hasConcept C38652104 @default.
- W2990147108 hasConcept C40305131 @default.
- W2990147108 hasConcept C41008148 @default.
- W2990147108 hasConcept C50644808 @default.
- W2990147108 hasConceptScore W2990147108C108583219 @default.
- W2990147108 hasConceptScore W2990147108C119857082 @default.
- W2990147108 hasConceptScore W2990147108C124101348 @default.
- W2990147108 hasConceptScore W2990147108C154945302 @default.
- W2990147108 hasConceptScore W2990147108C23130292 @default.
- W2990147108 hasConceptScore W2990147108C38652104 @default.
- W2990147108 hasConceptScore W2990147108C40305131 @default.
- W2990147108 hasConceptScore W2990147108C41008148 @default.
- W2990147108 hasConceptScore W2990147108C50644808 @default.
- W2990147108 hasLocation W29901471081 @default.
- W2990147108 hasOpenAccess W2990147108 @default.
- W2990147108 hasPrimaryLocation W29901471081 @default.
- W2990147108 hasRelatedWork W2922457425 @default.
- W2990147108 hasRelatedWork W3014300295 @default.
- W2990147108 hasRelatedWork W3124051732 @default.
- W2990147108 hasRelatedWork W3164822677 @default.
- W2990147108 hasRelatedWork W3215138031 @default.
- W2990147108 hasRelatedWork W4223943233 @default.
- W2990147108 hasRelatedWork W4225161397 @default.
- W2990147108 hasRelatedWork W4250304930 @default.
- W2990147108 hasRelatedWork W4299487748 @default.
- W2990147108 hasRelatedWork W4309045103 @default.
- W2990147108 isParatext "false" @default.
- W2990147108 isRetracted "false" @default.
- W2990147108 magId "2990147108" @default.
- W2990147108 workType "article" @default.