Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287864740> ?p ?o ?g. }
Showing items 1 to 74 of
74
with 100 items per page.
- W4287864740 abstract "Adversarial training has shown its ability in producing models that are robust to perturbations on the input data, but usually at the expense of decrease in the standard accuracy. To mitigate this issue, it is commonly believed that more training data will eventually help such adversarially robust models generalize better on the benign/unperturbed test data. In this paper, however, we challenge this conventional belief and show that more training data can hurt the generalization of adversarially robust models in the classification problems. We first investigate the Gaussian mixture classification with a linear loss and identify three regimes based on the strength of the adversary. In the weak adversary regime, more data improves the generalization of adversarially robust models. In the medium adversary regime, with more training data, the generalization loss exhibits a double descent curve, which implies the existence of an intermediate stage where more training data hurts the generalization. In the strong adversary regime, more data almost immediately causes the generalization error to increase. Then we move to the analysis of a two-dimensional classification problem with a 0-1 loss. We prove that more data always hurts the generalization performance of adversarially trained models with large perturbations. To complement our theoretical results, we conduct empirical studies on Gaussian mixture classification, support vector machines (SVMs), and linear regression." @default.
- W4287864740 created "2022-07-26" @default.
- W4287864740 creator A5020548562 @default.
- W4287864740 creator A5065562247 @default.
- W4287864740 creator A5078143614 @default.
- W4287864740 date "2020-02-25" @default.
- W4287864740 modified "2023-09-23" @default.
- W4287864740 title "The Curious Case of Adversarially Robust Models: More Data Can Help, Double Descend, or Hurt Generalization" @default.
- W4287864740 hasPublicationYear "2020" @default.
- W4287864740 type Work @default.
- W4287864740 citedByCount "0" @default.
- W4287864740 crossrefType "posted-content" @default.
- W4287864740 hasAuthorship W4287864740A5020548562 @default.
- W4287864740 hasAuthorship W4287864740A5065562247 @default.
- W4287864740 hasAuthorship W4287864740A5078143614 @default.
- W4287864740 hasBestOaLocation W42878647401 @default.
- W4287864740 hasConcept C104317684 @default.
- W4287864740 hasConcept C112313634 @default.
- W4287864740 hasConcept C11413529 @default.
- W4287864740 hasConcept C121332964 @default.
- W4287864740 hasConcept C12267149 @default.
- W4287864740 hasConcept C127716648 @default.
- W4287864740 hasConcept C134306372 @default.
- W4287864740 hasConcept C154945302 @default.
- W4287864740 hasConcept C163716315 @default.
- W4287864740 hasConcept C177148314 @default.
- W4287864740 hasConcept C185592680 @default.
- W4287864740 hasConcept C188082640 @default.
- W4287864740 hasConcept C21080849 @default.
- W4287864740 hasConcept C33923547 @default.
- W4287864740 hasConcept C37736160 @default.
- W4287864740 hasConcept C38652104 @default.
- W4287864740 hasConcept C41008148 @default.
- W4287864740 hasConcept C41065033 @default.
- W4287864740 hasConcept C51632099 @default.
- W4287864740 hasConcept C55493867 @default.
- W4287864740 hasConcept C62520636 @default.
- W4287864740 hasConceptScore W4287864740C104317684 @default.
- W4287864740 hasConceptScore W4287864740C112313634 @default.
- W4287864740 hasConceptScore W4287864740C11413529 @default.
- W4287864740 hasConceptScore W4287864740C121332964 @default.
- W4287864740 hasConceptScore W4287864740C12267149 @default.
- W4287864740 hasConceptScore W4287864740C127716648 @default.
- W4287864740 hasConceptScore W4287864740C134306372 @default.
- W4287864740 hasConceptScore W4287864740C154945302 @default.
- W4287864740 hasConceptScore W4287864740C163716315 @default.
- W4287864740 hasConceptScore W4287864740C177148314 @default.
- W4287864740 hasConceptScore W4287864740C185592680 @default.
- W4287864740 hasConceptScore W4287864740C188082640 @default.
- W4287864740 hasConceptScore W4287864740C21080849 @default.
- W4287864740 hasConceptScore W4287864740C33923547 @default.
- W4287864740 hasConceptScore W4287864740C37736160 @default.
- W4287864740 hasConceptScore W4287864740C38652104 @default.
- W4287864740 hasConceptScore W4287864740C41008148 @default.
- W4287864740 hasConceptScore W4287864740C41065033 @default.
- W4287864740 hasConceptScore W4287864740C51632099 @default.
- W4287864740 hasConceptScore W4287864740C55493867 @default.
- W4287864740 hasConceptScore W4287864740C62520636 @default.
- W4287864740 hasLocation W42878647401 @default.
- W4287864740 hasOpenAccess W4287864740 @default.
- W4287864740 hasPrimaryLocation W42878647401 @default.
- W4287864740 hasRelatedWork W10648265 @default.
- W4287864740 hasRelatedWork W13034104 @default.
- W4287864740 hasRelatedWork W1594625 @default.
- W4287864740 hasRelatedWork W3030276 @default.
- W4287864740 hasRelatedWork W5979161 @default.
- W4287864740 hasRelatedWork W6229082 @default.
- W4287864740 hasRelatedWork W6533109 @default.
- W4287864740 hasRelatedWork W7655147 @default.
- W4287864740 hasRelatedWork W8394581 @default.
- W4287864740 hasRelatedWork W5323488 @default.
- W4287864740 isParatext "false" @default.
- W4287864740 isRetracted "false" @default.
- W4287864740 workType "article" @default.