Matches in SemOpenAlex for { <https://semopenalex.org/work/W3154580345> ?p ?o ?g. }
- W3154580345 abstract "Lipschitz constrained models have been used to solve specifics deep learning problems such as the estimation of Wasserstein distance for GAN, or the training of neural networks robust to adversarial attacks. Regardless the novel and effective algorithms to build such 1-Lipschitz networks, their usage remains marginal, and they are commonly considered as less expressive and less able to fit properly the data than their unconstrained counterpart. The goal of the paper is to demonstrate that, despite being empirically harder to train, 1-Lipschitz neural networks are theoretically better grounded than unconstrained ones when it comes to classification. To achieve that we recall some results about 1-Lipschitz function in the scope of deep learning and we extend and illustrate them to derive general properties for classification. First, we show that 1-Lipschitz neural network can fit arbitrarily difficult frontier making them as expressive as classical ones. When minimizing the log loss, we prove that the optimization problem under Lipschitz constraint is well posed and have a minimum, whereas regular neural networks can diverge even on remarkably simple situations. Then, we study the link between classification with 1-Lipschitz network and optimal transport thanks to regularized versions of Kantorovich-Rubinstein duality theory. Last, we derive preliminary bounds on their VC dimension." @default.
- W3154580345 created "2021-04-26" @default.
- W3154580345 creator A5004249005 @default.
- W3154580345 creator A5032520414 @default.
- W3154580345 creator A5052743324 @default.
- W3154580345 creator A5060903205 @default.
- W3154580345 date "2021-04-11" @default.
- W3154580345 modified "2023-09-27" @default.
- W3154580345 title "The Many Faces of 1-Lipschitz Neural Networks." @default.
- W3154580345 cites W1495302720 @default.
- W3154580345 cites W1663973292 @default.
- W3154580345 cites W1804110266 @default.
- W3154580345 cites W2010353172 @default.
- W3154580345 cites W2019363670 @default.
- W3154580345 cites W2029538739 @default.
- W3154580345 cites W2054276177 @default.
- W3154580345 cites W2116261113 @default.
- W3154580345 cites W2119451546 @default.
- W3154580345 cites W2154952480 @default.
- W3154580345 cites W2156909104 @default.
- W3154580345 cites W2163641240 @default.
- W3154580345 cites W2194775991 @default.
- W3154580345 cites W2529714286 @default.
- W3154580345 cites W2618017917 @default.
- W3154580345 cites W2739748921 @default.
- W3154580345 cites W2800415562 @default.
- W3154580345 cites W2911634294 @default.
- W3154580345 cites W2962756933 @default.
- W3154580345 cites W2962847566 @default.
- W3154580345 cites W2962851953 @default.
- W3154580345 cites W2962879692 @default.
- W3154580345 cites W2963178695 @default.
- W3154580345 cites W2963212650 @default.
- W3154580345 cites W2963440492 @default.
- W3154580345 cites W2963466683 @default.
- W3154580345 cites W2963543570 @default.
- W3154580345 cites W2963556353 @default.
- W3154580345 cites W2963664410 @default.
- W3154580345 cites W2963685250 @default.
- W3154580345 cites W2963836885 @default.
- W3154580345 cites W2964116600 @default.
- W3154580345 cites W2964153729 @default.
- W3154580345 cites W2964253222 @default.
- W3154580345 cites W2970615870 @default.
- W3154580345 cites W2995623499 @default.
- W3154580345 cites W3016506115 @default.
- W3154580345 cites W3033357972 @default.
- W3154580345 cites W3034222740 @default.
- W3154580345 cites W3035154336 @default.
- W3154580345 cites W3101700548 @default.
- W3154580345 cites W3118608800 @default.
- W3154580345 cites W3120383865 @default.
- W3154580345 cites W3126829708 @default.
- W3154580345 cites W3130826642 @default.
- W3154580345 cites W3135117229 @default.
- W3154580345 cites W3159888837 @default.
- W3154580345 cites W3173448244 @default.
- W3154580345 hasPublicationYear "2021" @default.
- W3154580345 type Work @default.
- W3154580345 sameAs 3154580345 @default.
- W3154580345 citedByCount "0" @default.
- W3154580345 crossrefType "posted-content" @default.
- W3154580345 hasAuthorship W3154580345A5004249005 @default.
- W3154580345 hasAuthorship W3154580345A5032520414 @default.
- W3154580345 hasAuthorship W3154580345A5052743324 @default.
- W3154580345 hasAuthorship W3154580345A5060903205 @default.
- W3154580345 hasConcept C111472728 @default.
- W3154580345 hasConcept C118615104 @default.
- W3154580345 hasConcept C126255220 @default.
- W3154580345 hasConcept C138885662 @default.
- W3154580345 hasConcept C14036430 @default.
- W3154580345 hasConcept C154945302 @default.
- W3154580345 hasConcept C202444582 @default.
- W3154580345 hasConcept C22324862 @default.
- W3154580345 hasConcept C2524010 @default.
- W3154580345 hasConcept C2776036281 @default.
- W3154580345 hasConcept C2778023678 @default.
- W3154580345 hasConcept C2780586882 @default.
- W3154580345 hasConcept C33676613 @default.
- W3154580345 hasConcept C33923547 @default.
- W3154580345 hasConcept C41008148 @default.
- W3154580345 hasConcept C50644808 @default.
- W3154580345 hasConcept C78458016 @default.
- W3154580345 hasConcept C86803240 @default.
- W3154580345 hasConceptScore W3154580345C111472728 @default.
- W3154580345 hasConceptScore W3154580345C118615104 @default.
- W3154580345 hasConceptScore W3154580345C126255220 @default.
- W3154580345 hasConceptScore W3154580345C138885662 @default.
- W3154580345 hasConceptScore W3154580345C14036430 @default.
- W3154580345 hasConceptScore W3154580345C154945302 @default.
- W3154580345 hasConceptScore W3154580345C202444582 @default.
- W3154580345 hasConceptScore W3154580345C22324862 @default.
- W3154580345 hasConceptScore W3154580345C2524010 @default.
- W3154580345 hasConceptScore W3154580345C2776036281 @default.
- W3154580345 hasConceptScore W3154580345C2778023678 @default.
- W3154580345 hasConceptScore W3154580345C2780586882 @default.
- W3154580345 hasConceptScore W3154580345C33676613 @default.
- W3154580345 hasConceptScore W3154580345C33923547 @default.
- W3154580345 hasConceptScore W3154580345C41008148 @default.
- W3154580345 hasConceptScore W3154580345C50644808 @default.