Matches in SemOpenAlex for { <https://semopenalex.org/work/W3204489147> ?p ?o ?g. }
- W3204489147 abstract "We introduce Noisy Feature Mixup (NFM), an inexpensive yet effective method for data augmentation that combines the best of interpolation based training and noise injection schemes. Rather than training with convex combinations of pairs of examples and their labels, we use noise-perturbed convex combinations of pairs of data points in both input and feature space. This method includes mixup and manifold mixup as special cases, but it has additional advantages, including better smoothing of decision boundaries and enabling improved model robustness. We provide theory to understand this as well as the implicit regularization effects of NFM. Our theory is supported by empirical results, demonstrating the advantage of NFM, as compared to mixup and manifold mixup. We show that residual networks and vision transformers trained with NFM have favorable trade-offs between predictive accuracy on clean data and robustness with respect to various types of data perturbation across a range of computer vision benchmark datasets." @default.
- W3204489147 created "2021-10-11" @default.
- W3204489147 creator A5007032334 @default.
- W3204489147 creator A5028046389 @default.
- W3204489147 creator A5033006662 @default.
- W3204489147 creator A5055366983 @default.
- W3204489147 creator A5083533730 @default.
- W3204489147 date "2021-10-05" @default.
- W3204489147 modified "2023-09-27" @default.
- W3204489147 title "Noisy Feature Mixup" @default.
- W3204489147 cites W1945616565 @default.
- W3204489147 cites W2027595342 @default.
- W3204489147 cites W2032916024 @default.
- W3204489147 cites W2101278992 @default.
- W3204489147 cites W2108598243 @default.
- W3204489147 cites W2111296615 @default.
- W3204489147 cites W2111406701 @default.
- W3204489147 cites W2124136621 @default.
- W3204489147 cites W2132424367 @default.
- W3204489147 cites W2151239833 @default.
- W3204489147 cites W2153417333 @default.
- W3204489147 cites W2156909104 @default.
- W3204489147 cites W2159737176 @default.
- W3204489147 cites W2163605009 @default.
- W3204489147 cites W2302255633 @default.
- W3204489147 cites W2401231614 @default.
- W3204489147 cites W2460937040 @default.
- W3204489147 cites W2508156266 @default.
- W3204489147 cites W2529714286 @default.
- W3204489147 cites W2618235498 @default.
- W3204489147 cites W2640329709 @default.
- W3204489147 cites W2765407302 @default.
- W3204489147 cites W2786622092 @default.
- W3204489147 cites W2793731154 @default.
- W3204489147 cites W2799032899 @default.
- W3204489147 cites W2809895662 @default.
- W3204489147 cites W2887603965 @default.
- W3204489147 cites W2893562797 @default.
- W3204489147 cites W2896292273 @default.
- W3204489147 cites W2897355816 @default.
- W3204489147 cites W2921861056 @default.
- W3204489147 cites W2944619585 @default.
- W3204489147 cites W2949979904 @default.
- W3204489147 cites W2954996726 @default.
- W3204489147 cites W2962729158 @default.
- W3204489147 cites W2963026800 @default.
- W3204489147 cites W2963060032 @default.
- W3204489147 cites W2963703360 @default.
- W3204489147 cites W2963952467 @default.
- W3204489147 cites W2964331576 @default.
- W3204489147 cites W2964582580 @default.
- W3204489147 cites W2965130685 @default.
- W3204489147 cites W2968009374 @default.
- W3204489147 cites W2969738732 @default.
- W3204489147 cites W2970121940 @default.
- W3204489147 cites W2976501124 @default.
- W3204489147 cites W2979484367 @default.
- W3204489147 cites W2987875759 @default.
- W3204489147 cites W2992308087 @default.
- W3204489147 cites W2994088087 @default.
- W3204489147 cites W3006935033 @default.
- W3204489147 cites W3007125775 @default.
- W3204489147 cites W3007863415 @default.
- W3204489147 cites W3034541035 @default.
- W3204489147 cites W3034670733 @default.
- W3204489147 cites W3035032188 @default.
- W3204489147 cites W3035224233 @default.
- W3204489147 cites W3035364336 @default.
- W3204489147 cites W3035388858 @default.
- W3204489147 cites W3041890115 @default.
- W3204489147 cites W3042135962 @default.
- W3204489147 cites W3042925324 @default.
- W3204489147 cites W3043300996 @default.
- W3204489147 cites W3103089845 @default.
- W3204489147 cites W3104484212 @default.
- W3204489147 cites W3118608800 @default.
- W3204489147 cites W3127713124 @default.
- W3204489147 cites W3132297750 @default.
- W3204489147 cites W3133309250 @default.
- W3204489147 cites W3134621603 @default.
- W3204489147 cites W3142085127 @default.
- W3204489147 cites W3159778524 @default.
- W3204489147 cites W3160683869 @default.
- W3204489147 cites W3161120562 @default.
- W3204489147 cites W3169455327 @default.
- W3204489147 cites W3175116353 @default.
- W3204489147 cites W3175126267 @default.
- W3204489147 cites W2963459876 @default.
- W3204489147 cites W3091886042 @default.
- W3204489147 cites W3117555893 @default.
- W3204489147 hasPublicationYear "2021" @default.
- W3204489147 type Work @default.
- W3204489147 sameAs 3204489147 @default.
- W3204489147 citedByCount "0" @default.
- W3204489147 crossrefType "posted-content" @default.
- W3204489147 hasAuthorship W3204489147A5007032334 @default.
- W3204489147 hasAuthorship W3204489147A5028046389 @default.
- W3204489147 hasAuthorship W3204489147A5033006662 @default.
- W3204489147 hasAuthorship W3204489147A5055366983 @default.
- W3204489147 hasAuthorship W3204489147A5083533730 @default.