FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3204773485> ?p ?o ?g. }

• W3204773485 abstract "Convolutional Neural Networks (CNNs) have become the de facto gold standard in computer vision applications in the past years. Recently, however, new model architectures have been proposed challenging the status quo. The Vision Transformer (ViT) relies solely on attention modules, while the MLP-Mixer architecture substitutes the self-attention modules with Multi-Layer Perceptrons (MLPs). Despite their great success, CNNs have been widely known to be vulnerable to adversarial attacks, causing serious concerns for security-sensitive applications. Thus, it is critical for the community to know whether the newly proposed ViT and MLP-Mixer are also vulnerable to adversarial attacks. To this end, we empirically evaluate their adversarial robustness under several adversarial attack setups and benchmark them against the widely used CNNs. Overall, we find that the two architectures, especially ViT, are more robust than their CNN models. Using a toy example, we also provide empirical evidence that the lower adversarial robustness of CNNs can be partially attributed to their shift-invariant property. Our frequency analysis suggests that the most robust ViT architectures tend to rely more on low-frequency features compared with CNNs. Additionally, we have an intriguing finding that MLP-Mixer is extremely vulnerable to universal adversarial perturbations." @default.
• W3204773485 created "2021-10-11" @default.
• W3204773485 creator A5000590564 @default.
• W3204773485 creator A5012455275 @default.
• W3204773485 creator A5024432805 @default.
• W3204773485 creator A5057230698 @default.
• W3204773485 creator A5075652176 @default.
• W3204773485 date "2021-10-06" @default.
• W3204773485 modified "2023-09-27" @default.
• W3204773485 title "Adversarial Robustness Comparison of Vision Transformer and MLP-Mixer to CNNs" @default.
• W3204773485 cites W1912570122 @default.
• W3204773485 cites W2108598243 @default.
• W3204773485 cites W2194775991 @default.
• W3204773485 cites W2243397390 @default.
• W3204773485 cites W2543927648 @default.
• W3204773485 cites W2570685808 @default.
• W3204773485 cites W2603766943 @default.
• W3204773485 cites W2774644650 @default.
• W3204773485 cites W2913314773 @default.
• W3204773485 cites W2915002466 @default.
• W3204773485 cites W2930003153 @default.
• W3204773485 cites W2940744433 @default.
• W3204773485 cites W2943152387 @default.
• W3204773485 cites W2947469743 @default.
• W3204773485 cites W2950739196 @default.
• W3204773485 cites W2962700793 @default.
• W3204773485 cites W2962847335 @default.
• W3204773485 cites W2963060032 @default.
• W3204773485 cites W2963062382 @default.
• W3204773485 cites W2963070423 @default.
• W3204773485 cites W2963207607 @default.
• W3204773485 cites W2963341956 @default.
• W3204773485 cites W2963389226 @default.
• W3204773485 cites W2963403868 @default.
• W3204773485 cites W2963560523 @default.
• W3204773485 cites W2963595196 @default.
• W3204773485 cites W2963693747 @default.
• W3204773485 cites W2963703197 @default.
• W3204773485 cites W2963744840 @default.
• W3204773485 cites W2963857521 @default.
• W3204773485 cites W2964153729 @default.
• W3204773485 cites W2964205597 @default.
• W3204773485 cites W2964253222 @default.
• W3204773485 cites W2970115835 @default.
• W3204773485 cites W2970389371 @default.
• W3204773485 cites W2971028215 @default.
• W3204773485 cites W2994931756 @default.
• W3204773485 cites W2996851481 @default.
• W3204773485 cites W2998421476 @default.
• W3204773485 cites W3006406099 @default.
• W3204773485 cites W3006622081 @default.
• W3204773485 cites W3014153848 @default.
• W3204773485 cites W3018458867 @default.
• W3204773485 cites W3020482686 @default.
• W3204773485 cites W3034408878 @default.
• W3204773485 cites W3034445277 @default.
• W3204773485 cites W3034619610 @default.
• W3204773485 cites W3034643863 @default.
• W3204773485 cites W3088909400 @default.
• W3204773485 cites W3102103184 @default.
• W3204773485 cites W3106835084 @default.
• W3204773485 cites W3116489684 @default.
• W3204773485 cites W3117037102 @default.
• W3204773485 cites W3119786062 @default.
• W3204773485 cites W3121523901 @default.
• W3204773485 cites W3126337037 @default.
• W3204773485 cites W3127751679 @default.
• W3204773485 cites W3128723389 @default.
• W3204773485 cites W3130071011 @default.
• W3204773485 cites W3130987061 @default.
• W3204773485 cites W3131500599 @default.
• W3204773485 cites W3132890542 @default.
• W3204773485 cites W3133696297 @default.
• W3204773485 cites W3135696193 @default.
• W3204773485 cites W3138516171 @default.
• W3204773485 cites W3139049060 @default.
• W3204773485 cites W3139773203 @default.
• W3204773485 cites W3142085127 @default.
• W3204773485 cites W3143373604 @default.
• W3204773485 cites W3145185940 @default.
• W3204773485 cites W3157506437 @default.
• W3204773485 cites W3158360872 @default.
• W3204773485 cites W3161120562 @default.
• W3204773485 cites W3163203812 @default.
• W3204773485 cites W3164024107 @default.
• W3204773485 cites W3166513219 @default.
• W3204773485 cites W3171408944 @default.
• W3204773485 cites W3173053527 @default.
• W3204773485 cites W3175937066 @default.
• W3204773485 cites W3196621661 @default.
• W3204773485 hasPublicationYear "2021" @default.
• W3204773485 type Work @default.
• W3204773485 sameAs 3204773485 @default.
• W3204773485 citedByCount "0" @default.
• W3204773485 crossrefType "posted-content" @default.
• W3204773485 hasAuthorship W3204773485A5000590564 @default.
• W3204773485 hasAuthorship W3204773485A5012455275 @default.
• W3204773485 hasAuthorship W3204773485A5024432805 @default.
• W3204773485 hasAuthorship W3204773485A5057230698 @default.
• W3204773485 hasAuthorship W3204773485A5075652176 @default.

• next