Matches in SemOpenAlex for { <https://semopenalex.org/work/W4313442948> ?p ?o ?g. }
Showing items 1 to 70 of
70
with 100 items per page.
- W4313442948 abstract "Robust explanations of machine learning models are critical to establishing human trust in the models. The top-$k$ intersection is widely used to evaluate the robustness of explanations. However, most existing attacking and defense strategies are based on $ell_p$ norms, thus creating a mismatch between the evaluation and optimization objectives. To this end, we define explanation thickness for measuring top-$k$ salient features ranking stability, and design the textit{R2ET} algorithm based on a novel tractable surrogate to maximize the thickness and stabilize the top salient features efficiently. Theoretically, we prove a connection between R2ET and adversarial training; using a novel multi-objective optimization formulation and a generalization error bound, we further prove that the surrogate objective can improve both the numerical and statistical stability of the explanations. Experiments with a wide spectrum of network architectures and data modalities demonstrate that R2ET attains higher explanation robustness under stealthy attacks while retaining model accuracy." @default.
- W4313442948 created "2023-01-06" @default.
- W4313442948 creator A5024785349 @default.
- W4313442948 creator A5056248574 @default.
- W4313442948 creator A5067421575 @default.
- W4313442948 creator A5070538645 @default.
- W4313442948 creator A5078953397 @default.
- W4313442948 date "2022-12-28" @default.
- W4313442948 modified "2023-09-25" @default.
- W4313442948 title "Provable Robust Saliency-based Explanations" @default.
- W4313442948 doi "https://doi.org/10.48550/arxiv.2212.14106" @default.
- W4313442948 hasPublicationYear "2022" @default.
- W4313442948 type Work @default.
- W4313442948 citedByCount "0" @default.
- W4313442948 crossrefType "posted-content" @default.
- W4313442948 hasAuthorship W4313442948A5024785349 @default.
- W4313442948 hasAuthorship W4313442948A5056248574 @default.
- W4313442948 hasAuthorship W4313442948A5067421575 @default.
- W4313442948 hasAuthorship W4313442948A5070538645 @default.
- W4313442948 hasAuthorship W4313442948A5078953397 @default.
- W4313442948 hasBestOaLocation W43134429481 @default.
- W4313442948 hasConcept C104317684 @default.
- W4313442948 hasConcept C112972136 @default.
- W4313442948 hasConcept C11413529 @default.
- W4313442948 hasConcept C117765406 @default.
- W4313442948 hasConcept C119857082 @default.
- W4313442948 hasConcept C126255220 @default.
- W4313442948 hasConcept C134306372 @default.
- W4313442948 hasConcept C154945302 @default.
- W4313442948 hasConcept C177148314 @default.
- W4313442948 hasConcept C185592680 @default.
- W4313442948 hasConcept C2780719617 @default.
- W4313442948 hasConcept C33923547 @default.
- W4313442948 hasConcept C37736160 @default.
- W4313442948 hasConcept C41008148 @default.
- W4313442948 hasConcept C55493867 @default.
- W4313442948 hasConcept C63479239 @default.
- W4313442948 hasConceptScore W4313442948C104317684 @default.
- W4313442948 hasConceptScore W4313442948C112972136 @default.
- W4313442948 hasConceptScore W4313442948C11413529 @default.
- W4313442948 hasConceptScore W4313442948C117765406 @default.
- W4313442948 hasConceptScore W4313442948C119857082 @default.
- W4313442948 hasConceptScore W4313442948C126255220 @default.
- W4313442948 hasConceptScore W4313442948C134306372 @default.
- W4313442948 hasConceptScore W4313442948C154945302 @default.
- W4313442948 hasConceptScore W4313442948C177148314 @default.
- W4313442948 hasConceptScore W4313442948C185592680 @default.
- W4313442948 hasConceptScore W4313442948C2780719617 @default.
- W4313442948 hasConceptScore W4313442948C33923547 @default.
- W4313442948 hasConceptScore W4313442948C37736160 @default.
- W4313442948 hasConceptScore W4313442948C41008148 @default.
- W4313442948 hasConceptScore W4313442948C55493867 @default.
- W4313442948 hasConceptScore W4313442948C63479239 @default.
- W4313442948 hasLocation W43134429481 @default.
- W4313442948 hasLocation W43134429482 @default.
- W4313442948 hasOpenAccess W4313442948 @default.
- W4313442948 hasPrimaryLocation W43134429481 @default.
- W4313442948 hasRelatedWork W2381411913 @default.
- W4313442948 hasRelatedWork W2992532659 @default.
- W4313442948 hasRelatedWork W3035729345 @default.
- W4313442948 hasRelatedWork W4225307033 @default.
- W4313442948 hasRelatedWork W4295850094 @default.
- W4313442948 hasRelatedWork W4311734044 @default.
- W4313442948 hasRelatedWork W4379162851 @default.
- W4313442948 hasRelatedWork W4379255972 @default.
- W4313442948 hasRelatedWork W4382763378 @default.
- W4313442948 hasRelatedWork W4383955378 @default.
- W4313442948 isParatext "false" @default.
- W4313442948 isRetracted "false" @default.
- W4313442948 workType "article" @default.