{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,10]],"date-time":"2026-04-10T10:03:32Z","timestamp":1775815412395,"version":"3.50.1"},"reference-count":44,"publisher":"IOP Publishing","issue":"3","license":[{"start":{"date-parts":[[2024,8,13]],"date-time":"2024-08-13T00:00:00Z","timestamp":1723507200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2024,8,13]],"date-time":"2024-08-13T00:00:00Z","timestamp":1723507200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"funder":[{"DOI":"10.13039\/501100002509","name":"Keimyung University","doi-asserted-by":"crossref","award":["2023"],"award-info":[{"award-number":["2023"]}],"id":[{"id":"10.13039\/501100002509","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2024,9,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    In fields requiring high accountability, it is necessary to understand how deep-learning models make decisions when analyzing the causes of image classification. Concept-based interpretation methods have recently been introduced to reveal the internal mechanisms of deep learning models using high-level concepts. However, such methods are constrained by a trade-off between accuracy and interpretability. For instance, in real-world environments, unlike in well-curated training data, the accurate prediction of expected concepts becomes a challenge owing to the various distortions and complexities introduced by different objects. To overcome this tradeoff, we propose concept graph embedding models (CGEM), reflecting the complex dependencies and structures among concepts through the learning of mutual directionalities. The concept graph convolutional neural network (Concept GCN), a downstream task of CGEM, differs from previous methods that solely determine the presence of concepts because it performs a final classification based on the relationships between con- cepts learned through graph embedding. This process endows the model with high resilience even in the presence of incorrect concepts. In addition, we utilize a deformable bipartite GCN for object- centric concept encoding in the earlier stages, which enhances the homogeneity of the concepts. The experimental results show that, based on deformable concept encoding, the CGEM mitigates the trade-off between task accuracy and interpretability. Moreover, it was confirmed that this approach allows the model to increase the resilience and interpretability while maintaining robustness against various real-world concept distortions and incorrect concept interventions. Our code is available at\n                    <jats:ext-link xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" ext-link-type=\"uri\" xlink:href=\"https:\/\/github.com\/jumpsnack\/cgem\">https:\/\/github.com\/jumpsnack\/cgem<\/jats:ext-link>\n                    .\n                  <\/jats:p>","DOI":"10.1088\/2632-2153\/ad6ad2","type":"journal-article","created":{"date-parts":[[2024,8,2]],"date-time":"2024-08-02T19:01:54Z","timestamp":1722625314000},"page":"035042","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":2,"title":["Concept graph embedding models for enhanced accuracy and interpretability"],"prefix":"10.1088","volume":"5","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-7452-3897","authenticated-orcid":true,"given":"Sangwon","family":"Kim","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7284-0768","authenticated-orcid":true,"given":"Byoung Chul","family":"Ko","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"266","published-online":{"date-parts":[[2024,8,13]]},"reference":[{"key":"mlstad6ad2bib1","first-page":"pp 770","article-title":"Deep residual learning for image recognition","author":"He","year":"2016"},{"key":"mlstad6ad2bib2","first-page":"pp 6105","article-title":"Efficientnet: rethinking model scaling for convolutional neural networks","author":"Tan","year":"2019"},{"key":"mlstad6ad2bib3","first-page":"pp 1","article-title":"An image is worth 16 words: transformers for image recognition at scale","author":"Dosovitskiy","year":"2021"},{"key":"mlstad6ad2bib4","first-page":"pp 10012","article-title":"Swin transformer: hierarchical vision transformer using shifted windows","author":"Liu","year":"2021"},{"key":"mlstad6ad2bib5","first-page":"pp 10265","article-title":"Cross-modal learning with 3D deformable attention for action recognition","author":"Kim","year":"2023"},{"key":"mlstad6ad2bib6","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2021.108365","article-title":"A polarization fusion network with geometric feature embedding for sar ship classification","volume":"123","author":"Zhang","year":"2022","journal-title":"Pattern Recognit."},{"key":"mlstad6ad2bib7","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/LGRS.2021.3119875","article-title":"Squeeze-and-excitation laplacian pyramid network with dual-polarization feature fusion for ship classification in sar images","volume":"19","author":"Zhang","year":"2022","journal-title":"IEEE Geosci. Remote Sens. Lett."},{"key":"mlstad6ad2bib8","doi-asserted-by":"publisher","first-page":"190","DOI":"10.1016\/j.isprsjprs.2021.10.010","article-title":"Balance learning for ship detection from synthetic aperture radar remote sensing imagery","volume":"182","author":"Zhang","year":"2021","journal-title":"ISPRS J. Photogramm. Remote Sens."},{"key":"mlstad6ad2bib9","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/LGRS.2022.3189961","article-title":"A mask attention interaction and scale enhancement network for sar ship instance segmentation","volume":"19","author":"Zhang","year":"2022","journal-title":"IEEE Geosci. Remote Sens. Lett."},{"key":"mlstad6ad2bib10","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/LGRS.2022.3166387","article-title":"A full-level context squeeze-and-excitation roi extractor for sar ship instance segmentation","volume":"19","author":"Zhang","year":"2022","journal-title":"IEEE Geosci. Remote Sens. Lett."},{"key":"mlstad6ad2bib11","first-page":"pp 63","article-title":"Layer-wise relevance propagation for neural networks with local renormalization layers","author":"Binder","year":"2016"},{"key":"mlstad6ad2bib12","first-page":"4765","article-title":"A unified approach to interpreting model predictions","author":"Lundberg","year":"2017"},{"key":"mlstad6ad2bib13","first-page":"pp 1135","article-title":"Why should i trust you?: explaining the predictions of any classifier","author":"Tulio Ribeiro","year":"2016"},{"key":"mlstad6ad2bib14","first-page":"pp 618","article-title":"Grad-cam: visual explanations from deep networks via gradient-based localization","author":"Selvaraju","year":"2017"},{"key":"mlstad6ad2bib15","first-page":"pp 11162","article-title":"ViT-NeT: interpretable vision transformers with neural tree decoder","author":"Kim","year":"2022"},{"key":"mlstad6ad2bib16","doi-asserted-by":"publisher","first-page":"2067","DOI":"10.1109\/TAI.2023.3312645","article-title":"Neural tree decoder for interpretation of vision transformers","volume":"5","author":"Kim","year":"2023","journal-title":"IEEE Trans. Artif. Intell."},{"key":"mlstad6ad2bib17","first-page":"pp 700","article-title":"Concept-based explanation for fine-grained images and its application in infectious keratitis classification","author":"Fang","year":"2020"},{"key":"mlstad6ad2bib18","first-page":"pp 2668","article-title":"Interpretability beyond feature attribution: quantitative testing with concept activation vectors (tcav)","author":"Kim","year":"2018"},{"key":"mlstad6ad2bib19","first-page":"pp 21400","article-title":"concept embedding models: beyond the accuracy-explainability trade-off","author":"Espinosa Zarlenga","year":"2022"},{"key":"mlstad6ad2bib20","first-page":"pp 1","article-title":"Towards automatic concept-based explanations","author":"Ghorbani","year":"2019"},{"key":"mlstad6ad2bib21","first-page":"pp 5338","article-title":"Concept bottleneck models","author":"Wei Koh","year":"2020"},{"key":"mlstad6ad2bib22","first-page":"pp 1","article-title":"Probabilistic concept bottleneck models","author":"Kim","year":"2023"},{"key":"mlstad6ad2bib23","first-page":"pp 1","article-title":"Post-hoc concept bottleneck models","author":"Yuksekgonul","year":"2023"},{"key":"mlstad6ad2bib24","first-page":"pp 1","article-title":"Learning to receive help: intervention-aware concept embedding models","author":"Espinosa Zarlenga","year":"2023"},{"key":"mlstad6ad2bib25","first-page":"pp 785","article-title":"XGBoost: a s calable tree boosting system","author":"Chen","year":"2016"},{"key":"mlstad6ad2bib26","first-page":"1","article-title":"Catboost: gradient boosting with categorical features support","author":"Veronika Dorogush","year":"2018"},{"key":"mlstad6ad2bib27","doi-asserted-by":"publisher","first-page":"2522","DOI":"10.1038\/s42256-019-0138-9","article-title":"From local explanations to global understanding with explainable ai for trees","volume":"2","author":"Lundberg","year":"2020","journal-title":"Nat. Mach. Intell."},{"key":"mlstad6ad2bib28","first-page":"pp 1","article-title":"NBDT: neural-backed decision tree","author":"Wan","year":"2021"},{"key":"mlstad6ad2bib29","first-page":"pp 839","article-title":"Grad-cam++: generalized gradient-based visual explanations for deep convolutional networks","author":"Chattopadhay","year":"2018"},{"key":"mlstad6ad2bib30","first-page":"pp 19187","article-title":"Language in a bottle: language model guided concept bottlenecks for interpretable image classification","author":"Yang","year":"2023"},{"key":"mlstad6ad2bib31","first-page":"pp 25037","article-title":"Text-to-concept (and back) via cross-model alignment","author":"Moayeri","year":"2023"},{"key":"mlstad6ad2bib32","first-page":"pp 10962","article-title":"Learning bottleneck concepts in image classification","author":"Wang","year":"2023"},{"key":"mlstad6ad2bib33","article-title":"The Caltech-ucsd birds-200-2011 dataset","author":"Wah","year":"2011"},{"key":"mlstad6ad2bib34","first-page":"pp 3730","article-title":"Deep learning face attributes in the wild","author":"Liu","year":"2015"},{"key":"mlstad6ad2bib35","first-page":"pp 1778","article-title":"Describing objects by their attributes","author":"Farhadi","year":"2009"},{"key":"mlstad6ad2bib36","doi-asserted-by":"publisher","first-page":"2251","DOI":"10.1109\/TPAMI.2018.2857768","article-title":"Zero-shot learning-a comprehensive evaluation of the good, the bad and the ugly","volume":"41","author":"Xian","year":"2018","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"mlstad6ad2bib37","first-page":"pp 590","article-title":"Chexpert: a large chest radiograph dataset with uncertainty labels and expert comparison","author":"Irvin","year":"2019"},{"key":"mlstad6ad2bib38","doi-asserted-by":"publisher","first-page":"1452","DOI":"10.1109\/TPAMI.2017.2723009","article-title":"Places: A 10 million image database for scene recognition","volume":"40","author":"Zhou","year":"2017","journal-title":"IEEE Trans. Pattern Analy. Mach. Intell."},{"key":"mlstad6ad2bib39","doi-asserted-by":"publisher","first-page":"6923","DOI":"10.1109\/TPAMI.2021.3074057","article-title":"Deepgcns: making gcns go as deep as cnns","volume":"45","author":"Li","year":"2023","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"mlstad6ad2bib40","first-page":"pp 1","article-title":"Inductive representation learning on large graphs","author":"Hamilton","year":"2017"},{"key":"mlstad6ad2bib41","article-title":"How powerful are graph neural networks?","author":"Xu","year":"2019"},{"key":"mlstad6ad2bib42","article-title":"Graph attention networks","author":"Velickovic","year":"2018"},{"key":"mlstad6ad2bib43","first-page":"pp 1","article-title":"Generative adversarial nets","author":"Goodfellow","year":"2014"},{"key":"mlstad6ad2bib44","first-page":"pp 6840","article-title":"Denoising diffusion probabilistic models","author":"Ho","year":"2020"}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,8,13]],"date-time":"2024-08-13T09:16:53Z","timestamp":1723540613000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ad6ad2"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,8,13]]},"references-count":44,"journal-issue":{"issue":"3","published-online":{"date-parts":[[2024,8,13]]},"published-print":{"date-parts":[[2024,9,1]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/ad6ad2","relation":{"has-review":[{"id-type":"doi","id":"10.1088\/2632-2153\/AD6AD2\/v1\/decision1","asserted-by":"object"},{"id-type":"doi","id":"10.1088\/2632-2153\/AD6AD2\/v2\/review1","asserted-by":"object"},{"id-type":"doi","id":"10.1088\/2632-2153\/AD6AD2\/v2\/response1","asserted-by":"object"},{"id-type":"doi","id":"10.1088\/2632-2153\/AD6AD2\/v2\/decision1","asserted-by":"object"},{"id-type":"doi","id":"10.1088\/2632-2153\/AD6AD2\/v1\/review1","asserted-by":"object"}]},"ISSN":["2632-2153"],"issn-type":[{"value":"2632-2153","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,8,13]]},"assertion":[{"value":"Concept graph embedding models for enhanced accuracy and interpretability","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2024 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2024-03-12","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2024-08-01","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2024-08-13","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}