Matches in SemOpenAlex for { <https://semopenalex.org/work/W3136258283> ?p ?o ?g. }
- W3136258283 abstract "Adversarial training is an effective method to train deep learning models that are resilient to norm-bounded perturbations, with the cost of nominal performance drop. While adversarial training appears to enhance the robustness and safety of a deep model deployed in open-world decision-critical applications, counterintuitively, it induces undesired behaviors in robot learning settings. In this paper, we show theoretically and experimentally that neural controllers obtained via adversarial training are subjected to three types of defects, namely transient, systematic, and conditional errors. We first generalize adversarial training to a safety-domain optimization scheme allowing for more generic specifications. We then prove that such a learning process tends to cause certain error profiles. We support our theoretical results by a thorough experimental safety analysis in a robot-learning task. Our results suggest that adversarial training is not yet ready for robot learning." @default.
- W3136258283 created "2021-03-29" @default.
- W3136258283 creator A5025684214 @default.
- W3136258283 creator A5034362557 @default.
- W3136258283 creator A5036588940 @default.
- W3136258283 creator A5066830185 @default.
- W3136258283 creator A5080555605 @default.
- W3136258283 date "2021-03-15" @default.
- W3136258283 modified "2023-09-27" @default.
- W3136258283 title "Adversarial Training is Not Ready for Robot Learning" @default.
- W3136258283 cites W202219663 @default.
- W3136258283 cites W2044339004 @default.
- W3136258283 cites W2055027199 @default.
- W3136258283 cites W2099471712 @default.
- W3136258283 cites W2117539524 @default.
- W3136258283 cites W2194775991 @default.
- W3136258283 cites W2766462876 @default.
- W3136258283 cites W2773115527 @default.
- W3136258283 cites W2786104118 @default.
- W3136258283 cites W2787708942 @default.
- W3136258283 cites W2797294217 @default.
- W3136258283 cites W2887603965 @default.
- W3136258283 cites W2898963688 @default.
- W3136258283 cites W2911634294 @default.
- W3136258283 cites W2924512726 @default.
- W3136258283 cites W2946956149 @default.
- W3136258283 cites W2950787360 @default.
- W3136258283 cites W2962729158 @default.
- W3136258283 cites W2962872506 @default.
- W3136258283 cites W2963207607 @default.
- W3136258283 cites W2963359064 @default.
- W3136258283 cites W2963389226 @default.
- W3136258283 cites W2963440492 @default.
- W3136258283 cites W2963496101 @default.
- W3136258283 cites W2963501948 @default.
- W3136258283 cites W2963564844 @default.
- W3136258283 cites W2963626025 @default.
- W3136258283 cites W2963775761 @default.
- W3136258283 cites W2963952467 @default.
- W3136258283 cites W2964116600 @default.
- W3136258283 cites W2964153729 @default.
- W3136258283 cites W2964159205 @default.
- W3136258283 cites W2964253222 @default.
- W3136258283 cites W2964833278 @default.
- W3136258283 cites W2967292964 @default.
- W3136258283 cites W2967762122 @default.
- W3136258283 cites W2967835402 @default.
- W3136258283 cites W2968352263 @default.
- W3136258283 cites W2970115835 @default.
- W3136258283 cites W2971109239 @default.
- W3136258283 cites W2972226851 @default.
- W3136258283 cites W2977505966 @default.
- W3136258283 cites W2980620660 @default.
- W3136258283 cites W2980728855 @default.
- W3136258283 cites W2989576588 @default.
- W3136258283 cites W2990408532 @default.
- W3136258283 cites W2991620912 @default.
- W3136258283 cites W2996325578 @default.
- W3136258283 cites W3003495965 @default.
- W3136258283 cites W3008151842 @default.
- W3136258283 cites W3008209689 @default.
- W3136258283 cites W3014152613 @default.
- W3136258283 cites W3028525609 @default.
- W3136258283 cites W3033739625 @default.
- W3136258283 cites W3041890115 @default.
- W3136258283 cites W3090132620 @default.
- W3136258283 cites W3090943552 @default.
- W3136258283 cites W3095242670 @default.
- W3136258283 cites W3096541186 @default.
- W3136258283 cites W3098018961 @default.
- W3136258283 cites W3112385784 @default.
- W3136258283 cites W3137837033 @default.
- W3136258283 cites W9657784 @default.
- W3136258283 cites W2740664948 @default.
- W3136258283 hasPublicationYear "2021" @default.
- W3136258283 type Work @default.
- W3136258283 sameAs 3136258283 @default.
- W3136258283 citedByCount "8" @default.
- W3136258283 countsByYear W31362582832021 @default.
- W3136258283 crossrefType "posted-content" @default.
- W3136258283 hasAuthorship W3136258283A5025684214 @default.
- W3136258283 hasAuthorship W3136258283A5034362557 @default.
- W3136258283 hasAuthorship W3136258283A5036588940 @default.
- W3136258283 hasAuthorship W3136258283A5066830185 @default.
- W3136258283 hasAuthorship W3136258283A5080555605 @default.
- W3136258283 hasConcept C104317684 @default.
- W3136258283 hasConcept C108583219 @default.
- W3136258283 hasConcept C119857082 @default.
- W3136258283 hasConcept C121332964 @default.
- W3136258283 hasConcept C134306372 @default.
- W3136258283 hasConcept C153294291 @default.
- W3136258283 hasConcept C154945302 @default.
- W3136258283 hasConcept C185592680 @default.
- W3136258283 hasConcept C2777211547 @default.
- W3136258283 hasConcept C33923547 @default.
- W3136258283 hasConcept C34388435 @default.
- W3136258283 hasConcept C37736160 @default.
- W3136258283 hasConcept C41008148 @default.
- W3136258283 hasConcept C50644808 @default.
- W3136258283 hasConcept C55493867 @default.