Current studentsExisting staffContact us
 

Improve interpretability of Information Bottlenecks for Attribution with Layer-wise Relevance Propagation

Simple item page
dc.contributor.authorChen, X.
dc.contributor.authorLi, J.
dc.contributor.authorLiu, J.
dc.contributor.authorPeters, S.
dc.contributor.authorLiu, L.
dc.contributor.authorLe, T.D.
dc.contributor.authorWalsh, A.
dc.contributor.conference2023 IEEE International Conference on Big Data (BigData) (15 Dec 2023 - 18 Dec 2023 : Sorrento, Italy)
dc.contributor.editorHe, J.
dc.date.issued2023
dc.description.abstractResearchers have developed various visualization techniques, such as attribution maps, to understand which parts of an input contribute most to a model’s decision. However, existing methods often produce disparate results and may lack human-perceptual interpretability. In this work, we propose Relevance-IBA, a novel approach that combines the strengths of Information Bottleneck Attribution (IBA) and Layer-wise Relevance Propagation’s (LRP) method to estimate more accurate and human-perceptually interpretable attribution maps. Our method accentuates the contours and subtle details of the identified object, making the model’s decisions more intuitively understandable. Additionally, we introduce a segmentation-oriented evaluation technique, which assesses the capacity of interpretability methods by emphasizing the most important pixels within an object’s boundaries. We benchmark Relevance-IBA against various methodologies, including DeepLIFT, Integrated Gradients, Guided-BP, Guided-GradCAM, IBA, and InputIBA. Our results indicate that Relevance-IBA not only boosts attribution accuracy but also prioritizes human-perceptual clarity, making it a valuable tool for interpreting complex model behaviors.
dc.identifier.citationProceedings 2023 IEEE International Conference on Big Data Bigdata 2023, 2023 / He, J. (ed./s), pp.1064-1069
dc.identifier.doi10.1109/BigData59044.2023.10386271
dc.identifier.isbn9798350324457
dc.identifier.orcidLiu, J. [0000-0002-0794-0404]
dc.identifier.orcidLe, T.D. [0000-0002-9732-4313]
dc.identifier.urihttps://hdl.handle.net/11541.2/37896
dc.language.isoen
dc.publisherIEEE
dc.publisher.placeUS
dc.rightsCopyright 2023 IEEE
dc.source.urihttps://doi.org/10.1109/bigdata59044.2023.10386271
dc.subjectIBA
dc.subjectattribution maps
dc.subjectinterpretability
dc.subjectLRP
dc.subjecthuman-perceptual
dc.titleImprove interpretability of Information Bottlenecks for Attribution with Layer-wise Relevance Propagation
dc.typeConference paper
pubs.publication-statusPublished
ror.mmsid9916831827901831

Files

Collections