{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,19]],"date-time":"2026-03-19T17:49:44Z","timestamp":1773942584517,"version":"3.50.1"},"reference-count":43,"publisher":"Wiley","issue":"3","license":[{"start":{"date-parts":[[2024,6,10]],"date-time":"2024-06-10T00:00:00Z","timestamp":1717977600000},"content-version":"vor","delay-in-days":9,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":["onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Computer Graphics Forum"],"published-print":{"date-parts":[[2024,6]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:p>We present concept\u2010aligned neurons, or CAN, a visualization design for comparing deep neural networks. The goal of CAN is to support users in understanding the similarities and differences between neural networks, with an emphasis on comparing neuron functionality across different models. To make this comparison intuitive, CAN uses concept\u2010based representations of neurons to visually align models in an interpretable manner. A key feature of CAN is the hierarchical organization of concepts, which permits users to relate sets of neurons at different levels of detail. CAN's visualization is designed to help compare the semantic coverage of neurons, as well as assess the distinctiveness, redundancy, and multi\u2010semantic alignment of neurons or groups of neurons, all at different concept granularity. We demonstrate the generality and effectiveness of CAN by comparing models trained on different datasets, neural networks with different architectures, and models trained for different objectives, e.g. adversarial robustness, and robustness to out\u2010of\u2010distribution data.<\/jats:p>","DOI":"10.1111\/cgf.15085","type":"journal-article","created":{"date-parts":[[2024,6,10]],"date-time":"2024-06-10T14:43:11Z","timestamp":1718030591000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":1,"title":["CAN: Concept\u2010Aligned Neurons for Visual Comparison of Deep Neural Network Models"],"prefix":"10.1111","volume":"43","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-0457-8091","authenticated-orcid":false,"given":"M.","family":"Li","sequence":"first","affiliation":[{"name":"Vanderbilt University  USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1626-7469","authenticated-orcid":false,"given":"S.","family":"Jeong","sequence":"additional","affiliation":[{"name":"Vanderbilt University  USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6455-8391","authenticated-orcid":false,"given":"S.","family":"Liu","sequence":"additional","affiliation":[{"name":"Lawrence Livermore National Laboratory  USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8876-2418","authenticated-orcid":false,"given":"M.","family":"Berger","sequence":"additional","affiliation":[{"name":"Vanderbilt University  USA"}]}],"member":"311","published-online":{"date-parts":[[2024,6,10]]},"reference":[{"key":"e_1_2_12_2_2","doi-asserted-by":"crossref","unstructured":"BoggustA. CarterB. SatyanarayanA.: Embedding comparator: Visualizing differences in global structure and local neighborhoods via small multiples. In27th international conference on intelligent user interfaces(2022) pp.746\u2013766. 3","DOI":"10.1145\/3490099.3511122"},{"key":"e_1_2_12_3_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2017.2744683"},{"key":"e_1_2_12_4_2","unstructured":"BykovK. KopfL. NakajimaS. KloftM. H\u00f6hneM. M.: Labeling neural representations with inverse recognition. InThirty\u2010seventh Conference on Neural Information Processing Systems(2023). 2 3 4 5"},{"key":"e_1_2_12_5_2","doi-asserted-by":"crossref","unstructured":"BauD. ZhouB. KhoslaA. OlivaA. TorralbaA.: Network dissection: Quantifying interpretability of deep visual representations. InProceedings of the IEEE conference on computer vision and pattern recognition(2017) pp.6541\u20136549. 2","DOI":"10.1109\/CVPR.2017.354"},{"key":"e_1_2_12_6_2","unstructured":"BauD. ZhuJ.\u2010Y. StrobeltH. ZhouB. TenenbaumJ. B. FreemanW. T. TorralbaA.: Gan dissection: Visualizing and understanding generative adversarial networks.arXiv preprint arXiv:1811.10597(2018). 2"},{"key":"e_1_2_12_7_2","unstructured":"DosovitskiyA. BeyerL. KolesnikovA. WeissenbornD. ZhaiX. UnterthinerT. DehghaniM. MindererM. HeigoldG. GellyS. et al.: An image is worth 16\u00d716 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929(2020). 7"},{"key":"e_1_2_12_8_2","first-page":"1","volume-title":"2017 26th international conference on computer communication and networks (ICCCN)","author":"Dodge S.","year":"2017"},{"key":"e_1_2_12_9_2","first-page":"7694","volume-title":"International Conference on Machine Learning","author":"Desai K.","year":"2023"},{"key":"e_1_2_12_10_2","doi-asserted-by":"crossref","unstructured":"FongR. VedaldiA.: Net2vec: Quantifying and explaining how concepts are encoded by filters in deep neural networks. InProceedings of the IEEE conference on computer vision and pattern recognition(2018) pp.8730\u20138738. 2","DOI":"10.1109\/CVPR.2018.00910"},{"key":"e_1_2_12_11_2","first-page":"181","volume-title":"Computer Graphics Forum","author":"Gleicher M.","year":"2020"},{"key":"e_1_2_12_12_2","doi-asserted-by":"crossref","unstructured":"G\u00f6rtlerJ. HohmanF. MoritzD. WongsuphasawatK. RenD. NairR. KirchnerM. PatelK.: Neo: Generalizing confusion matrix visualization to hierarchical and multi\u2010output labels. InProceedings of the 2022 CHI Conference on Human Factors in Computing Systems(2022) pp.1\u201313. 3","DOI":"10.1145\/3491102.3501823"},{"key":"e_1_2_12_13_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2017.2744199"},{"key":"e_1_2_12_14_2","unstructured":"GoodfellowI. J. ShlensJ. SzegedyC.: Explaining and harnessing adversarial examples.arXiv preprint arXiv:1412.6572(2014). 9"},{"key":"e_1_2_12_15_2","article-title":"Towards automatic concept\u2010based explanations","volume":"32","author":"Ghorbani A.","year":"2019","journal-title":"Advances in neural information processing systems"},{"issue":"1","key":"e_1_2_12_16_2","first-page":"74","article-title":"Visual concept programming: A visual analytics approach to injecting human intelligence at scale","volume":"29","author":"Hoque M. N.","year":"2022","journal-title":"IEEE Transactions on Visualization and Computer Graphics"},{"key":"e_1_2_12_17_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2020.3045918"},{"key":"e_1_2_12_18_2","unstructured":"HendrycksD. MazeikaM. DietterichT.: Deep anomaly detection with outlier exposure. InInternational Conference on Learning Representations(2018). 4"},{"issue":"1","key":"e_1_2_12_19_2","first-page":"831","article-title":"Conceptexplainer: Interactive explanation for deep neural networks from a concept perspective","volume":"29","author":"Huang J.","year":"2022","journal-title":"IEEE Transactions on Visualization and Computer Graphics"},{"key":"e_1_2_12_20_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2019.2934659"},{"key":"e_1_2_12_21_2","unstructured":"HernandezE. SchwettmannS. BauD. BagashviliT. TorralbaA. AndreasJ.: Natural language descriptions of deep visual features. InInternational Conference on Learning Representations(2021). 2 3 5"},{"key":"e_1_2_12_22_2","doi-asserted-by":"crossref","unstructured":"HeK. ZhangX. RenS. SunJ.: Deep residual learning for image recognition. InProceedings of the IEEE conference on computer vision and pattern recognition(2016) pp.770\u2013778. 4 7","DOI":"10.1109\/CVPR.2016.90"},{"key":"e_1_2_12_23_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2017.2744718"},{"key":"e_1_2_12_24_2","doi-asserted-by":"crossref","unstructured":"KielaD. BartoloM. NieY. KaushikD. GeigerA. WuZ. VidgenB. PrasadG. SinghA. RingshiaP. et al.: Dynabench: Rethinking benchmarking in nlp. InProceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies(2021) pp.4110\u20134124. 2","DOI":"10.18653\/v1\/2021.naacl-main.324"},{"key":"e_1_2_12_25_2","first-page":"3519","volume-title":"International conference on machine learning","author":"Kornblith S.","year":"2019"},{"key":"e_1_2_12_26_2","first-page":"2668","volume-title":"International conference on machine learning","author":"Kim B.","year":"2018"},{"key":"e_1_2_12_27_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2016.2598831"},{"key":"e_1_2_12_28_2","first-page":"17153","article-title":"Compositional explanations of neurons","volume":"33","author":"Mu J.","year":"2020","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_2_12_29_2","first-page":"728","volume-title":"European Conference on Computer Vision","author":"Minderer M.","year":"2022"},{"key":"e_1_2_12_30_2","doi-asserted-by":"publisher","DOI":"10.21105\/joss.00861"},{"key":"e_1_2_12_31_2","doi-asserted-by":"publisher","DOI":"10.1145\/219717.219748"},{"key":"e_1_2_12_31_3","doi-asserted-by":"crossref","unstructured":"doi:10.1145\/219717.219748. 2 5 7","DOI":"10.1145\/219717.219748"},{"key":"e_1_2_12_32_2","unstructured":"OikarinenT. WengT.\u2010W.: Clip\u2010dissect: Automatic description of neuron representations in deep vision networks.arXiv preprint arXiv:2204.10965(2022). 2 3 4 5 6"},{"key":"e_1_2_12_33_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2021.3114858"},{"key":"e_1_2_12_34_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2017.2744358"},{"key":"e_1_2_12_35_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2016.2598828"},{"key":"e_1_2_12_36_2","first-page":"8748","volume-title":"International conference on machine learning","author":"Radford A.","year":"2021"},{"key":"e_1_2_12_37_2","first-page":"12116","article-title":"Do vision transformers see like convolutional neural networks?","volume":"34","author":"Raghu M.","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_2_12_38_2","unstructured":"UtreraF. KravitzE. ErichsonN. B. KhannaR. MahoneyM. W.: Adversarially\u2010trained deep nets transfer better: Illustration on image classification. InInternational Conference on Learning Representations(2020). 4 7"},{"issue":"6","key":"e_1_2_12_39_2","first-page":"2326","article-title":"Vac\u2010cnn: A visual analytics system for comparative studies of deep convolutional neural networks","volume":"28","author":"Xuan X.","year":"2022","journal-title":"IEEE Transactions on Visualization and Computer Graphics"},{"key":"e_1_2_12_40_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10590-1_53"},{"key":"e_1_2_12_41_2","unstructured":"ZengH. HaleemH. PlantazX. CaoN. QuH.: Cnncomparator: Comparative analytics of convolutional neural networks.arXiv preprint arXiv:1710.05285(2017). 9"},{"key":"e_1_2_12_42_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2018.2864499"},{"key":"e_1_2_12_43_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2021.3114837"}],"container-title":["Computer Graphics Forum"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1111\/cgf.15085","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,6,14]],"date-time":"2024-06-14T12:26:41Z","timestamp":1718368001000},"score":1,"resource":{"primary":{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/10.1111\/cgf.15085"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,6]]},"references-count":43,"journal-issue":{"issue":"3","published-print":{"date-parts":[[2024,6]]}},"alternative-id":["10.1111\/cgf.15085"],"URL":"https:\/\/doi.org\/10.1111\/cgf.15085","archive":["Portico"],"relation":{},"ISSN":["0167-7055","1467-8659"],"issn-type":[{"value":"0167-7055","type":"print"},{"value":"1467-8659","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,6]]},"assertion":[{"value":"2024-06-10","order":2,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e15085"}}