Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386721366> ?p ?o ?g. }
Showing items 1 to 51 of
51
with 100 items per page.
- W4386721366 abstract "While enjoying the great achievements brought by deep learning (DL), people are also worried about the decision made by DL models, since the high degree of non-linearity of DL models makes the decision extremely difficult to understand. Consequently, attacks such as adversarial attacks are easy to carry out, but difficult to detect and explain, which has led to a boom in the research on local explanation methods for explaining model decisions. In this paper, we evaluate the faithfulness of explanation methods and find that traditional tests on faithfulness encounter the random dominance problem, ie, the random selection performs the best, especially for complex data. To further solve this problem, we propose three trend-based faithfulness tests and empirically demonstrate that the new trend tests can better assess faithfulness than traditional tests on image, natural language and security tasks. We implement the assessment system and evaluate ten popular explanation methods. Benefiting from the trend tests, we successfully assess the explanation methods on complex data for the first time, bringing unprecedented discoveries and inspiring future research. Downstream tasks also greatly benefit from the tests. For example, model debugging equipped with faithful explanation methods performs much better for detecting and correcting accuracy and security problems." @default.
- W4386721366 created "2023-09-14" @default.
- W4386721366 creator A5017417068 @default.
- W4386721366 creator A5021024679 @default.
- W4386721366 creator A5041680615 @default.
- W4386721366 creator A5049302472 @default.
- W4386721366 creator A5088713944 @default.
- W4386721366 date "2023-09-09" @default.
- W4386721366 modified "2023-09-30" @default.
- W4386721366 title "Good-looking but Lacking Faithfulness: Understanding Local Explanation Methods through Trend-based Testing" @default.
- W4386721366 doi "https://doi.org/10.1145/3576915.3616605" @default.
- W4386721366 hasPublicationYear "2023" @default.
- W4386721366 type Work @default.
- W4386721366 citedByCount "0" @default.
- W4386721366 crossrefType "posted-content" @default.
- W4386721366 hasAuthorship W4386721366A5017417068 @default.
- W4386721366 hasAuthorship W4386721366A5021024679 @default.
- W4386721366 hasAuthorship W4386721366A5041680615 @default.
- W4386721366 hasAuthorship W4386721366A5049302472 @default.
- W4386721366 hasAuthorship W4386721366A5088713944 @default.
- W4386721366 hasBestOaLocation W43867213661 @default.
- W4386721366 hasConcept C119857082 @default.
- W4386721366 hasConcept C154945302 @default.
- W4386721366 hasConcept C168065819 @default.
- W4386721366 hasConcept C199360897 @default.
- W4386721366 hasConcept C37736160 @default.
- W4386721366 hasConcept C41008148 @default.
- W4386721366 hasConcept C81917197 @default.
- W4386721366 hasConceptScore W4386721366C119857082 @default.
- W4386721366 hasConceptScore W4386721366C154945302 @default.
- W4386721366 hasConceptScore W4386721366C168065819 @default.
- W4386721366 hasConceptScore W4386721366C199360897 @default.
- W4386721366 hasConceptScore W4386721366C37736160 @default.
- W4386721366 hasConceptScore W4386721366C41008148 @default.
- W4386721366 hasConceptScore W4386721366C81917197 @default.
- W4386721366 hasLocation W43867213661 @default.
- W4386721366 hasOpenAccess W4386721366 @default.
- W4386721366 hasPrimaryLocation W43867213661 @default.
- W4386721366 hasRelatedWork W1483845062 @default.
- W4386721366 hasRelatedWork W1522854984 @default.
- W4386721366 hasRelatedWork W1578053891 @default.
- W4386721366 hasRelatedWork W1601811574 @default.
- W4386721366 hasRelatedWork W1602801198 @default.
- W4386721366 hasRelatedWork W1987935534 @default.
- W4386721366 hasRelatedWork W2120071210 @default.
- W4386721366 hasRelatedWork W3046843850 @default.
- W4386721366 hasRelatedWork W4386716251 @default.
- W4386721366 hasRelatedWork W97732546 @default.
- W4386721366 isParatext "false" @default.
- W4386721366 isRetracted "false" @default.
- W4386721366 workType "article" @default.