Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035897475> ?p ?o ?g. }
- W3035897475 abstract "Despite neural networks have achieved prominent performance on many natural language processing (NLP) tasks, they are vulnerable to adversarial examples. In this paper, we propose Dirichlet Neighborhood Ensemble (DNE), a randomized smoothing method for training a robust model to defense substitution-based attacks. During training, DNE forms virtual sentences by sampling embedding vectors for each word in an input sentence from a convex hull spanned by the word and its synonyms, and it augments them with the training data. In such a way, the model is robust to adversarial attacks while maintaining the performance on the original clean data. DNE is agnostic to the network architectures and scales to large models for NLP applications. We demonstrate through extensive experimentation that our method consistently outperforms recently proposed defense methods by a significant margin across different network architectures and multiple data sets." @default.
- W3035897475 created "2020-06-25" @default.
- W3035897475 creator A5010841999 @default.
- W3035897475 creator A5017835517 @default.
- W3035897475 creator A5066608665 @default.
- W3035897475 creator A5087096372 @default.
- W3035897475 creator A5088834359 @default.
- W3035897475 date "2020-06-20" @default.
- W3035897475 modified "2023-09-27" @default.
- W3035897475 title "Defense against Adversarial Attacks in NLP via Dirichlet Neighborhood Ensemble." @default.
- W3035897475 cites W1539309091 @default.
- W3035897475 cites W1840435438 @default.
- W3035897475 cites W2113459411 @default.
- W3035897475 cites W2243397390 @default.
- W3035897475 cites W2250539671 @default.
- W3035897475 cites W2413794162 @default.
- W3035897475 cites W2735135478 @default.
- W3035897475 cites W2759471388 @default.
- W3035897475 cites W2767962654 @default.
- W3035897475 cites W2772621923 @default.
- W3035897475 cites W2798384383 @default.
- W3035897475 cites W2799194071 @default.
- W3035897475 cites W2803392236 @default.
- W3035897475 cites W2898963688 @default.
- W3035897475 cites W2911634294 @default.
- W3035897475 cites W2922293812 @default.
- W3035897475 cites W2926587947 @default.
- W3035897475 cites W2947251152 @default.
- W3035897475 cites W2949128310 @default.
- W3035897475 cites W2962710014 @default.
- W3035897475 cites W2962713901 @default.
- W3035897475 cites W2962718684 @default.
- W3035897475 cites W2962943487 @default.
- W3035897475 cites W2963012544 @default.
- W3035897475 cites W2963026768 @default.
- W3035897475 cites W2963083752 @default.
- W3035897475 cites W2963207607 @default.
- W3035897475 cites W2963341956 @default.
- W3035897475 cites W2963857521 @default.
- W3035897475 cites W2963952467 @default.
- W3035897475 cites W2963969878 @default.
- W3035897475 cites W2964082701 @default.
- W3035897475 cites W2964121744 @default.
- W3035897475 cites W2964153729 @default.
- W3035897475 cites W2964232431 @default.
- W3035897475 cites W2964253222 @default.
- W3035897475 cites W2970078867 @default.
- W3035897475 cites W2970449623 @default.
- W3035897475 cites W2980248764 @default.
- W3035897475 cites W2995368830 @default.
- W3035897475 cites W2995876497 @default.
- W3035897475 cites W2998277219 @default.
- W3035897475 cites W3035422387 @default.
- W3035897475 cites W3036267641 @default.
- W3035897475 cites W3093419064 @default.
- W3035897475 hasPublicationYear "2020" @default.
- W3035897475 type Work @default.
- W3035897475 sameAs 3035897475 @default.
- W3035897475 citedByCount "15" @default.
- W3035897475 countsByYear W30358974752020 @default.
- W3035897475 countsByYear W30358974752021 @default.
- W3035897475 crossrefType "posted-content" @default.
- W3035897475 hasAuthorship W3035897475A5010841999 @default.
- W3035897475 hasAuthorship W3035897475A5017835517 @default.
- W3035897475 hasAuthorship W3035897475A5066608665 @default.
- W3035897475 hasAuthorship W3035897475A5087096372 @default.
- W3035897475 hasAuthorship W3035897475A5088834359 @default.
- W3035897475 hasConcept C119857082 @default.
- W3035897475 hasConcept C154945302 @default.
- W3035897475 hasConcept C204321447 @default.
- W3035897475 hasConcept C2524010 @default.
- W3035897475 hasConcept C2777462759 @default.
- W3035897475 hasConcept C2777530160 @default.
- W3035897475 hasConcept C31972630 @default.
- W3035897475 hasConcept C33923547 @default.
- W3035897475 hasConcept C3770464 @default.
- W3035897475 hasConcept C37736160 @default.
- W3035897475 hasConcept C41008148 @default.
- W3035897475 hasConcept C41608201 @default.
- W3035897475 hasConcept C774472 @default.
- W3035897475 hasConcept C90805587 @default.
- W3035897475 hasConceptScore W3035897475C119857082 @default.
- W3035897475 hasConceptScore W3035897475C154945302 @default.
- W3035897475 hasConceptScore W3035897475C204321447 @default.
- W3035897475 hasConceptScore W3035897475C2524010 @default.
- W3035897475 hasConceptScore W3035897475C2777462759 @default.
- W3035897475 hasConceptScore W3035897475C2777530160 @default.
- W3035897475 hasConceptScore W3035897475C31972630 @default.
- W3035897475 hasConceptScore W3035897475C33923547 @default.
- W3035897475 hasConceptScore W3035897475C3770464 @default.
- W3035897475 hasConceptScore W3035897475C37736160 @default.
- W3035897475 hasConceptScore W3035897475C41008148 @default.
- W3035897475 hasConceptScore W3035897475C41608201 @default.
- W3035897475 hasConceptScore W3035897475C774472 @default.
- W3035897475 hasConceptScore W3035897475C90805587 @default.
- W3035897475 hasLocation W30358974751 @default.
- W3035897475 hasOpenAccess W3035897475 @default.
- W3035897475 hasPrimaryLocation W30358974751 @default.
- W3035897475 hasRelatedWork W2113459411 @default.
- W3035897475 hasRelatedWork W2250539671 @default.