GF-LRP: A Method for Explaining Predictions Made by Variational Graph Auto-Encoders
Variational graph autoencoders (VGAEs) combine the best of graph convolutional networks (GCNs) and variational inference and have been used to address various tasks such as node classification or link prediction. However, the lack of explainability is a limiting factor when trustworthy decisions are...
Uloženo v:
| Vydáno v: | IEEE transactions on emerging topics in computational intelligence Ročník 9; číslo 1; s. 281 - 291 |
|---|---|
| Hlavní autoři: | , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
Piscataway
IEEE
01.02.2025
The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Témata: | |
| ISSN: | 2471-285X, 2471-285X |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Abstract | Variational graph autoencoders (VGAEs) combine the best of graph convolutional networks (GCNs) and variational inference and have been used to address various tasks such as node classification or link prediction. However, the lack of explainability is a limiting factor when trustworthy decisions are required. In this paper, we present a novel post-hoc explainability framework for VGAEs that considers their encoder-decoder architecture. Specifically, we propose a layer-wise-relevance-propagation-based (LRP-based) explanation technique coined GF-LRP which, to our knowledge, is the first explanation method for VGAEs. GF-LRP goes beyond existing LRP techniques for GCNs by taking into account, in addition to input features and the graph structure of the data, the VGAE branch-specific architecture. The explanations are branch-specific in the sense that we explain the mean and standard deviation branches of the Gaussian distribution learned by the model. For a node's prediction, GF-LRP infers the most relevant features, nodes and its edges. To prove the effectiveness of our explanation method, we compute fidelity, sparsity and contrastivity as well as commonly employed evaluation metrics. Extensive experiments and visualizations on two real-world datasets demonstrate the effectiveness of the proposed explanation method. |
|---|---|
| AbstractList | Variational graph autoencoders (VGAEs) combine the best of graph convolutional networks (GCNs) and variational inference and have been used to address various tasks such as node classification or link prediction. However, the lack of explainability is a limiting factor when trustworthy decisions are required. In this paper, we present a novel post-hoc explainability framework for VGAEs that considers their encoder-decoder architecture. Specifically, we propose a layer-wise-relevance-propagation-based (LRP-based) explanation technique coined GF-LRP which, to our knowledge, is the first explanation method for VGAEs. GF-LRP goes beyond existing LRP techniques for GCNs by taking into account, in addition to input features and the graph structure of the data, the VGAE branch-specific architecture. The explanations are branch-specific in the sense that we explain the mean and standard deviation branches of the Gaussian distribution learned by the model. For a node's prediction, GF-LRP infers the most relevant features, nodes and its edges. To prove the effectiveness of our explanation method, we compute fidelity, sparsity and contrastivity as well as commonly employed evaluation metrics. Extensive experiments and visualizations on two real-world datasets demonstrate the effectiveness of the proposed explanation method. |
| Author | Deligiannis, Nikos Rodrigo-Bonet, Esther |
| Author_xml | – sequence: 1 givenname: Esther orcidid: 0000-0003-4129-3295 surname: Rodrigo-Bonet fullname: Rodrigo-Bonet, Esther email: erodrigo@etrovub.be organization: Department of Electronics and Informatics, Vrije Universiteit Brussel, Brussels, Belgium – sequence: 2 givenname: Nikos orcidid: 0000-0001-9300-5860 surname: Deligiannis fullname: Deligiannis, Nikos email: ndeligia@etrovub.be organization: Department of Electronics and Informatics, Vrije Universiteit Brussel, Brussels, Belgium |
| BookMark | eNpNkE9PAjEQxRuDiYh8AeOhiefFttvZP94IASSBSBSNt6bbnZUluF3bJYFv7yIcmMtMJu-9vPxuSaeyFRJyz9mAc5Y-rcar0WwgmJCDUPI05vKKdIWMeSAS-Opc3Dek7_2GMSZS4CHILnmfToL52_KZDukCm7XNaWEdHe_rrS6rsvqmS4d5aZrSVp4udI40O9BP7Up9fOktnTpdr-lw19hgXBmbo_N35LrQW4_98-6Rj0lb8SWYv05no-E8MG2hJjAGCgghRp2mGnMQmYyTKIeo0DKBImqHZVyDSFguTKozyVCCziApMOSShz3yeMqtnf3doW_Uxu5cW8qrkEMKLOSJbFXipDLOeu-wULUrf7Q7KM7UkZ_656eO_NSZX2t6OJlKRLwwQBLFbe4f6jdsUg |
| CODEN | ITETCU |
| Cites_doi | 10.1371/journal.pone.0181142 10.18653/v1/2021.naacl-main.333 10.1109/TNN.2008.2005605 10.1109/ACCESS.2020.3018033 10.1016/j.patcog.2023.109874 10.1109/TPAMI.2022.3204236 10.1007/978-3-319-10590-1_53 10.1109/ICASSP.2019.8683787 10.5555/2969033.2969125 10.1109/CNS.2019.8802833 10.1038/nature14539 10.1371/journal.pone.0130140 10.1109/JIOT.2020.2999446 10.1109/TPAMI.2021.3116668 10.1109/CVPR.2019.01103 10.1109/TKDE.2022.3187455 10.1007/978-3-030-28954-6_10 10.1109/TPAMI.2021.3115452 10.1145/3359786 10.18653/v1/W17-5221 10.1007/978-3-030-28954-6_16 10.1109/TSIPN.2022.3180679 10.1145/3236009 10.1145/3331184.3331267 10.1613/jair.1.12228 10.1007/978-3-030-57321-8_4 10.1109/tnnls.2024.3370918 10.1561/2200000056 10.18653/v1/N16-3020 10.1016/S0304-3800(02)00257-0 10.1186/s13059-023-02850-y 10.1117/12.2511964 10.1109/CVPR42600.2020.00867 |
| ContentType | Journal Article |
| Copyright | Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2025 |
| Copyright_xml | – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2025 |
| DBID | 97E RIA RIE AAYXX CITATION 7SP 8FD L7M |
| DOI | 10.1109/TETCI.2024.3419714 |
| DatabaseName | IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE Electronic Library (IEL) CrossRef Electronics & Communications Abstracts Technology Research Database Advanced Technologies Database with Aerospace |
| DatabaseTitle | CrossRef Technology Research Database Advanced Technologies Database with Aerospace Electronics & Communications Abstracts |
| DatabaseTitleList | Technology Research Database |
| Database_xml | – sequence: 1 dbid: RIE name: IEEE Electronic Library (IEL) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Architecture |
| EISSN | 2471-285X |
| EndPage | 291 |
| ExternalDocumentID | 10_1109_TETCI_2024_3419714 10586750 |
| Genre | orig-research |
| GrantInformation_xml | – fundername: Interuniversitair Micro-Electronica Centrum VZW; IMEC funderid: 10.13039/100020844 – fundername: Ph.D. Fellowship Strategic Basic Research grantid: 1SC4521N – fundername: Vlaamse regering; Flemish Government funderid: 10.13039/501100011878 – fundername: Onderzoeksprogramma Artificiële Intelligentie (AI) Vlaanderen" Programme – fundername: Research Foundation - Flanders – fundername: AAA Project |
| GroupedDBID | 0R~ 97E AAJGR AASAJ AAWTH ABJNI ABQJQ ABVLG ACGFS AGQYO AHBIQ AKJIK AKQYR ALMA_UNASSIGNED_HOLDINGS ATWAV BEFXN BFFAM BGNUA BKEBE BPEOZ EBS EJD IFIPE JAVBF OCL RIA RIE AAYXX CITATION 7SP 8FD ABAZT L7M |
| ID | FETCH-LOGICAL-c247t-cc5f5357ea99aed52b4786d56fa485f66660b1a5280d2c9ab40e45ab58fe31413 |
| IEDL.DBID | RIE |
| ISICitedReferencesCount | 2 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001263417500001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 2471-285X |
| IngestDate | Mon Jun 30 12:58:08 EDT 2025 Sat Nov 29 05:12:11 EST 2025 Wed Dec 10 09:49:59 EST 2025 |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 1 |
| Language | English |
| License | https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html https://doi.org/10.15223/policy-029 https://doi.org/10.15223/policy-037 |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c247t-cc5f5357ea99aed52b4786d56fa485f66660b1a5280d2c9ab40e45ab58fe31413 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 |
| ORCID | 0000-0003-4129-3295 0000-0001-9300-5860 |
| PQID | 3159503184 |
| PQPubID | 4437216 |
| PageCount | 11 |
| ParticipantIDs | proquest_journals_3159503184 crossref_primary_10_1109_TETCI_2024_3419714 ieee_primary_10586750 |
| PublicationCentury | 2000 |
| PublicationDate | 2025-02-01 |
| PublicationDateYYYYMMDD | 2025-02-01 |
| PublicationDate_xml | – month: 02 year: 2025 text: 2025-02-01 day: 01 |
| PublicationDecade | 2020 |
| PublicationPlace | Piscataway |
| PublicationPlace_xml | – name: Piscataway |
| PublicationTitle | IEEE transactions on emerging topics in computational intelligence |
| PublicationTitleAbbrev | TETCI |
| PublicationYear | 2025 |
| Publisher | IEEE The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| Publisher_xml | – name: IEEE – name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE) |
| References | ref13 ref34 ref15 Springenberg (ref35) 2015 ref14 ref36 ref31 Ying (ref21) 2019 ref30 ref33 ref10 ref32 ref2 ref1 ref17 ref39 ref16 Baldassarre (ref24) 2019 Kipf (ref28) 2017 ref38 ref19 ref18 Shrikumar (ref37) 2017 Luo (ref20) 2020 Kipf (ref26) 2016 ref23 ref45 ref25 ref42 ref41 ref22 ref44 ref43 ref27 ref29 Gautam (ref5) 2022 ref8 ref7 ref9 ref4 ref3 Adel (ref6) 2018 Draizen (ref11) 2022 ref40 Han (ref12) 2022 |
| References_xml | – ident: ref15 doi: 10.1371/journal.pone.0181142 – ident: ref30 doi: 10.18653/v1/2021.naacl-main.333 – ident: ref27 doi: 10.1109/TNN.2008.2005605 – ident: ref29 doi: 10.1109/ACCESS.2020.3018033 – ident: ref41 doi: 10.1016/j.patcog.2023.109874 – ident: ref19 doi: 10.1109/TPAMI.2022.3204236 – ident: ref38 doi: 10.1007/978-3-319-10590-1_53 – ident: ref43 doi: 10.1109/ICASSP.2019.8683787 – ident: ref2 doi: 10.5555/2969033.2969125 – ident: ref39 doi: 10.5555/2969033.2969125 – ident: ref4 doi: 10.1109/CNS.2019.8802833 – volume-title: Proc. Int. Conf. Learn. Representations Workshop Track year: 2015 ident: ref35 article-title: Striving for simplicity: The all convolutional net – volume-title: Proc. Int. Conf. Learn. Representations year: 2017 ident: ref28 article-title: Semi-supervised classification with graph convolutional networks – ident: ref1 doi: 10.1038/nature14539 – ident: ref14 doi: 10.1371/journal.pone.0130140 – ident: ref31 doi: 10.1109/JIOT.2020.2999446 – ident: ref42 doi: 10.1109/TPAMI.2021.3116668 – start-page: 17940 volume-title: Proc. Int. Conf. Neural Inf. Process. Syst. year: 2022 ident: ref5 article-title: ProtoVAE: A trustworthy self-explainable prototypical variational model – start-page: 19620 volume-title: Proc. Int. Conf. Neural Inf. Process. Syst. year: 2020 ident: ref20 article-title: Parameterized explainer for graph neural network – ident: ref23 doi: 10.1109/CVPR.2019.01103 – ident: ref22 doi: 10.1109/TKDE.2022.3187455 – ident: ref10 doi: 10.1007/978-3-030-28954-6_10 – ident: ref18 doi: 10.1109/TPAMI.2021.3115452 – volume-title: Proc. ICML Workshop Learn. Reasoning Graph-Structured Representations year: 2019 ident: ref24 article-title: Explainability techniques for graph convolutional networks – start-page: 32 volume-title: Proc. Int. Conf. Neural Inf. Process. Syst. year: 2019 ident: ref21 article-title: GNNExplainer: Generating explanations for graph neural networks – ident: ref8 doi: 10.1145/3359786 – ident: ref16 doi: 10.18653/v1/W17-5221 – ident: ref17 doi: 10.1007/978-3-030-28954-6_16 – start-page: 5256 volume-title: Proc. Int. Conf. Neural Inf. Process. Syst. year: 2022 ident: ref12 article-title: Which explanation should I choose? A function approximation perspective to characterizing post hoc explanations – ident: ref45 doi: 10.1109/TSIPN.2022.3180679 – ident: ref32 doi: 10.1145/3236009 – ident: ref44 doi: 10.1145/3331184.3331267 – ident: ref25 doi: 10.1613/jair.1.12228 – ident: ref34 doi: 10.1007/978-3-030-57321-8_4 – ident: ref40 doi: 10.1109/tnnls.2024.3370918 – year: 2022 ident: ref11 article-title: Explainable deep generative models, ancestral fragments, and murky regions of the protein structure universe publication-title: bioRxiv – ident: ref3 doi: 10.1561/2200000056 – volume-title: Proc. Int. Conf. Neural Inf. Process. Syst. year: 2016 ident: ref26 article-title: Variational graph auto-encoders – start-page: 50 volume-title: Proc. 35th Int. Conf. Mach. Learn. year: 2018 ident: ref6 article-title: Discovering interpretable representations for both deep generative and discriminative models – ident: ref33 doi: 10.18653/v1/N16-3020 – start-page: 3145 volume-title: Proc. Int. Conf. Learn. Representations year: 2017 ident: ref37 article-title: Learning important features through propagating activation differences – ident: ref36 doi: 10.1016/S0304-3800(02)00257-0 – ident: ref7 doi: 10.1186/s13059-023-02850-y – ident: ref9 doi: 10.1117/12.2511964 – ident: ref13 doi: 10.1109/CVPR42600.2020.00867 |
| SSID | ssj0002951354 |
| Score | 2.295312 |
| Snippet | Variational graph autoencoders (VGAEs) combine the best of graph convolutional networks (GCNs) and variational inference and have been used to address various... |
| SourceID | proquest crossref ieee |
| SourceType | Aggregation Database Index Database Publisher |
| StartPage | 281 |
| SubjectTerms | Architecture Artificial neural networks Autoencoders Computational modeling Data models Deep learning Effectiveness Encoders-Decoders Explainable deep learning geometric deep learning Graph convolutional networks graph convolutional neural networks Graph neural networks Graph theory layer-wise relevance propagation Network analysis Normal distribution post-hoc explanations Predictive models variational autoencoders |
| Title | GF-LRP: A Method for Explaining Predictions Made by Variational Graph Auto-Encoders |
| URI | https://ieeexplore.ieee.org/document/10586750 https://www.proquest.com/docview/3159503184 |
| Volume | 9 |
| WOSCitedRecordID | wos001263417500001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| journalDatabaseRights | – providerCode: PRVIEE databaseName: IEEE Electronic Library (IEL) customDbUrl: eissn: 2471-285X dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0002951354 issn: 2471-285X databaseCode: RIE dateStart: 20170101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8NAEF60eBDBZ8VqlT14k63pZjebeAvSVsGWolV6C_sKeEklTQX_vbObVAriwVsOCWxmdna-mZ2ZD6Hr3AhqOOUkV8wQFltGFLeG9JVmVGsNCJV5sgkxmcTzeTJtmtV9L4y11hef2Z579Hf5ZqFXLlUGFs5jALgQoW8LEdXNWj8JFQpYIeRs3RgTJLezwez-EUJAynpuapnv1NlwPp5N5dcR7P3K8OCfKzpE-w2AxGmt8SO0ZYtjtJdu3AecoJfRkDw9T-9wiseeIRoDNMWu3K7mg8DT0t3P-C2Hx9JYrL7wG0TNTWYQj9wYa5yuqgUZFK7rvVy20esQ_vSBNOwJRFMmKgJyznnIhZVJIi3oQzERR4ZHuWQxzyFsiQLVl5zGgaE6kYoFlnGpeJzbsA--7RS1ikVhzxAWmgptAC0kbj5gX0gGdk-NSZh2xamyg27WYs0-6iEZmQ8ugiTzSsicErJGCR3UdoLceLOWYQd116rIGkNaZiHALe4OHnb-x2cXaJc6Tl5fSd1Frapc2Uu0oz-r92V55ffINzONuZg |
| linkProvider | IEEE |
| linkToHtml | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8NAEF6kCorgs2K16h68ydZkM9sk3oL0hW0pWqW3sK-Al1bSVPDfu7tJpSAevOWQwGZmZ-eb2Zn5ELrNVEgVo4xkAhSBSAMRTCviCwlUSmkQKjiyiXA8jmazeFI1q7teGK21Kz7TLfvo7vLVQq5sqsxYOIsMwDUR-jYDoF7ZrvWTUqEGLQQM1q0xXnw_7UwfByYIpNCyc8tcr86G-3F8Kr8OYedZuof_XNMROqggJE5KnR-jLT0_QfvJxo3AKXrpdcnwefKAEzxyHNHYgFNsC-5KRgg8ye0Njdt0eMSVxuILv5m4ucoN4p4dZI2TVbEgnbnte8-XdfTaNX_aJxV_ApEUwoIYSWcsYKHmccy10YiAMGor1s44RCwzgUvbEz5nNPIUlTEX4GlgXLAo04FvvNsZqs0Xc32OcChpKJXBC7GdEOiHHIzlU6VikLY8lTfQ3Vqs6Uc5JiN14YUXp04JqVVCWimhgepWkBtvljJsoOZaFWllSss0MICL2aMHLv747Abt9qejYTocjJ8u0R61DL2urrqJakW-0ldoR34W78v82u2XbwopvN8 |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=GF-LRP%3A+A+Method+for+Explaining+Predictions+Made+by+Variational+Graph+Auto-Encoders&rft.jtitle=IEEE+transactions+on+emerging+topics+in+computational+intelligence&rft.au=Rodrigo-Bonet%2C+Esther&rft.au=Deligiannis%2C+Nikos&rft.date=2025-02-01&rft.issn=2471-285X&rft.eissn=2471-285X&rft.volume=9&rft.issue=1&rft.spage=281&rft.epage=291&rft_id=info:doi/10.1109%2FTETCI.2024.3419714&rft.externalDBID=n%2Fa&rft.externalDocID=10_1109_TETCI_2024_3419714 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2471-285X&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2471-285X&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2471-285X&client=summon |