{"type":"ScholarlyArticle","title":"Metadata-driven Task Relation Discovery for Multi-task Learning","authors":[{"name":"Zimu Zheng","slug":"zimu-zheng","given":"Zimu","family":"Zheng","affiliation":"The Hong Kong Polytechnic University , Huawei Technologies Co.Ltd."},{"name":"Yuqi Wang","slug":"yuqi-wang","given":"Yuqi","family":"Wang","affiliation":"The Hong Kong Polytechnic University , Fujian Nebula Big Data Application Service Co. , Ltd."},{"name":"Quanyu Dai","slug":"quanyu-dai","given":"Quanyu","family":"Dai","affiliation":"The Hong Kong Polytechnic University"},{"name":"Huadi Zheng","slug":"huadi-zheng","given":"Huadi","family":"Zheng","affiliation":"The Hong Kong Polytechnic University"},{"name":"Dan Wang","slug":"dan-wang","given":"Dan","family":"Wang","affiliation":"The Hong Kong Polytechnic University"}],"abstract":"Task Relation Discovery (TRD), i.e., reveal the relation of tasks, has notable value: it is the key concept underlying Multi-task Learning (MTL) and provides a principled way for identifying redundancies across tasks. However, task relation is usually specifically determined by data scientist resulting in the additional human effort for TRD, while transfer based on brute-force methods or mere training samples may cause negative effects which degrade the learning performance. To avoid negative transfer in an automatic manner, our idea is to leverage commonly available context attributes in nowadays systems, i.e., the metadata. In this paper, we, for the first time, introduce metadata into TRD for MTL and propose a novel Metadata Clustering method, which jointly uses historical samples and additional metadata to automatically exploit the true relatedness. It also avoids the negative transfer by identifying reusable samples between related tasks. Experimental results on five real-world datasets demonstrate that the proposed method is effective for MTL with TRD, and particularly useful in complicated systems with diverse metadata but insufficient data samples. In general, this study helps in automatic relation discovery among partially related tasks and sheds new light on the development of TRD in MTL through the use of metadata as apriori information.","doi":"10.24963/ijcai.2019/615","url":"https://ijcai.org/proceedings/2019/615","pdf":"https://ijcai.org/proceedings/2019/0615.pdf","markdown":"https://ijcai.org/proceedings/2019/615.md","bibtex":"https://ijcai.org/proceedings/2019/bibtex/615","volume":{"slug":"2019","title":"Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2019","url":"https://ijcai.org/proceedings/2019/","year":2019,"location":"Macao","dates":"10-16 August 2019","editors":"Sarit Kraus","isbn":"978-0-9992411-4-1","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2019"},"section":"Machine Learning M-Z","track":"Main track","number":615,"pages":{"first":4426,"last":4432},"keywords":["Machine Learning: Transfer, Adaptation, Multi-task Learning","Machine Learning Applications: Networks","Machine Learning: Clustering","Machine Learning Applications: Other Applications","Multidisciplinary Topics and Applications: Ubiquitous Computing Systems"],"videos":[],"references":[{"position":1,"raw":"[Ando, 2006] Rie Kubota Ando. Applying alternating struc- ture optimization to word sense disambiguation. In CoNLL, pages 77-84, 2006.","title":"Applying alternating structure optimization to word sense disambiguation","authors":"Rie Kubota, Ando","year":"2006","venue":"CoNLL","doi":null,"url":null,"internal_url":null},{"position":2,"raw":"[Australian Institute of Health and Welfare, 2018] Australian Institute of Health and Welfare. Project meteor: Metadata repository. https://meteor.aihw.gov.au/ content/index.phtml/itemId/181162, 2018.","title":"Project meteor: Metadata repository","authors":null,"year":"2018","venue":null,"doi":null,"url":"https://meteor.aihw.gov.au/content/index.phtml/itemId/181162,2018","internal_url":null},{"position":3,"raw":"[Balaji and others, 2016] Bharathan Balaji et al. Brick: To- wards a unified metadata schema for buildings. In Pro- ceedings of the 3rd ACM BuildSys, pages 41-50, 2016.","title":"Brick: Towards a unified metadata schema for buildings","authors":"Bharathan Balaji","year":"2016","venue":"Proceedings of the 3rd ACM BuildSys","doi":null,"url":null,"internal_url":null},{"position":4,"raw":"[Baxter and others, 2000] Jonathan Baxter et al. A model of inductive bias learning. J. Artif. Intell. Res.(JAIR), 2000. [BETS Research Group, Accessed 1 DEC 2017] BETS Re- search Group. A uniform metadata schema for buildings. https://brickschema.org/, Accessed: 1 DEC 2017.","title":"A model of inductive bias learning","authors":"Jonathan Baxter","year":"2000","venue":"BETS Research Group. A uniform metadata schema for buildings","doi":null,"url":"https://brickschema.org/","internal_url":null},{"position":5,"raw":"[Caruana, 1997] Rich Caruana. Multitask learning. Machine Learning, 28(1):41-75, 1997.","title":"Multitask learning","authors":"Caruana","year":"1997","venue":"Machine Learning","doi":null,"url":null,"internal_url":null},{"position":6,"raw":"[Choi and others, 2017] Edward Choi et al. Gram: Graph- based attention model for healthcare representation learn- ing. In ACM SIGKDD, pages 787-795, 2017.","title":"Gram: Graphbased attention model for healthcare representation learning","authors":"Edward Choi, Choi","year":"2017","venue":"ACM SIGKDD","doi":null,"url":null,"internal_url":null},{"position":7,"raw":"[de Roux and others, 2018] Daniel de Roux et al. Tax fraud detection for under-reporting declarations using an unsu- pervised machine learning approach. In ACM SIGKDD, pages 215-222, 2018.","title":"Tax fraud detection for under-reporting declarations using an unsupervised machine learning approach","authors":"Daniel De, Roux","year":"2018","venue":"ACM SIGKDD","doi":null,"url":null,"internal_url":null},{"position":8,"raw":"[Dunstan and Humphrey, 2005] Tim Dunstan and Charles Humphrey. Discovering microdata variables. In Statistics Canada, pages No. 11-522-XIE, 2005.","title":"Discovering microdata variables","authors":"Dunstan, Humphrey ;, Tim Dunstan, Charles Humphrey","year":"2005","venue":"Statistics Canada","doi":null,"url":null,"internal_url":null},{"position":9,"raw":"[Evgeniou and others, 2005] Theodoros Evgeniou et al. Learning multiple tasks with kernel methods. JMLR, 6(Apr):615-637, 2005.","title":"Learning multiple tasks with kernel methods","authors":"Theodoros Evgeniou","year":"2005","venue":"JMLR","doi":null,"url":null,"internal_url":null},{"position":10,"raw":"[Evgeniou and Pontil, 2004] Theodoros Evgeniou and Mas- similiano Pontil. Regularized multi-task learning. In ACM SIGKDD, pages 109-117, 2004.","title":"Theodoros Evgeniou and Massimiliano Pontil. Regularized multi-task learning","authors":"Pontil Evgeniou","year":"2004","venue":"ACM SIGKDD","doi":null,"url":null,"internal_url":null},{"position":11,"raw":"[Gower, 1971] John C Gower. A general coefficient of sim- ilarity and some of its properties. Biometrics, pages 857- 871, 1971.","title":"A general coefficient of similarity and some of its properties","authors":"C John, Gower","year":"1971","venue":"Biometrics","doi":null,"url":null,"internal_url":null},{"position":12,"raw":"[Gu and Zhou, 2009] Quanquan Gu and Jie Zhou. Learning the shared subspace for multi-task clustering and transduc- tive transfer classification. In IEEE ICDM, 2009.","title":"Learning the shared subspace for multi-task clustering and transductive transfer classification","authors":"Zhou ; ] Quanquan Gu, Jie Gu, Zhou","year":"2009","venue":"IEEE ICDM","doi":null,"url":null,"internal_url":null},{"position":13,"raw":"[Han and others, 2014] Lei Han et al. Encoding tree spar- sity in multi-task learning: A probabilistic framework. In AAAI, pages 1854-1860, 2014.","title":"Encoding tree sparsity in multi-task learning: A probabilistic framework","authors":"Others ; Lei Han, Han","year":"2014","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":14,"raw":"[Isele et al., 2016] David Isele, Mohammad Rostami, and Eric Eaton. Using task features for zero-shot knowledge transfer in lifelong learning. In IJCAI, 2016.","title":"Using task features for zero-shot knowledge transfer in lifelong learning","authors":"David Isele, Mohammad Rostami, Eric Eaton","year":"2016","venue":"IJCAI","doi":null,"url":null,"internal_url":"/proceedings/2016/232"},{"position":15,"raw":"[ISO/ IEC JTC 1, 2018] ISO/ IEC JTC 1. Iso/iec 11179, information technology -metadata registries. http:// metadata-standards.org/11179/, 2018.","title":"information technology -metadata registries","authors":null,"year":"2018","venue":null,"doi":null,"url":"http://metadata-standards.org/11179/,2018","internal_url":null},{"position":16,"raw":"[Jaccard, 1908] Paul Jaccard. Nouvelles researches sur la distribution florale. Bull Soc Vaud Sci Nat, 44, 1908.","title":"Nouvelles researches sur la distribution florale","authors":"Paul Jaccard","year":"1908","venue":"Bull Soc Vaud Sci Nat","doi":null,"url":null,"internal_url":null},{"position":17,"raw":"[Kato et al., 2008] Tsuyoshi Kato, Hisashi Kashima, Masashi Sugiyama, and Kiyoshi Asai. Multi-task learning via conic programming. In NIPS, pages 737-744, 2008.","title":"Multi-task learning via conic programming","authors":"Tsuyoshi Kato, Hisashi Kashima, Masashi Sugiyama, Kiyoshi Asai","year":"2008","venue":"NIPS","doi":null,"url":null,"internal_url":null},{"position":18,"raw":"[Larochelle et al., 2008] Hugo Larochelle, Dumitru Erhan, and Yoshua Bengio. Zero-data learning of new tasks. In AAAI, volume 1, page 3, 2008.","title":"Zero-data learning of new tasks","authors":"Larochelle","year":"2008","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":19,"raw":"[Legendre and Legendre, 2012] Pierre Legendre and Loic FJ Legendre. Numerical ecology, volume 24. Elsevier, 2012.","title":"Pierre Legendre and Loic FJ Legendre","authors":"Legendre Legendre","year":"2012","venue":null,"doi":null,"url":null,"internal_url":null},{"position":20,"raw":"[Liu and others, 2007] Ying Liu et al. Tableseer: automatic table metadata extraction and searching in digital libraries. In ACM/IEEE JCDL, pages 91-100, 2007.","title":"Tableseer: automatic table metadata extraction and searching in digital libraries","authors":"Ying Liu","year":"2007","venue":"ACM/IEEE JCDL","doi":null,"url":null,"internal_url":null},{"position":21,"raw":"[Liu and others, 2017] An-An Liu et al. Hierarchical clus- tering multi-task learning for joint human action grouping and recognition. IEEE TPAMI, 39(1):102-114, 2017.","title":"An-An Liu et al. Hierarchical clustering multi-task learning for joint human action grouping and recognition","authors":"Others Liu","year":"2017","venue":"IEEE TPAMI","doi":null,"url":null,"internal_url":null},{"position":22,"raw":"[Pan and Yang, 2010] Sinno Jialin Pan and Qiang Yang. A survey on transfer learning. IEEE Transactions on knowl- edge and data engineering, 22(10):1345-1359, 2010.","title":"A survey on transfer learning","authors":"Jialin Sinno, Qiang Pan, Yang","year":"2010","venue":"IEEE Transactions on knowledge and data engineering","doi":null,"url":null,"internal_url":null},{"position":23,"raw":"[Podani, 1999] János Podani. Extending gower's general co- efficient of similarity to ordinal characters. Taxon, pages 331-340, 1999.","title":"János Podani. Extending gower's general coefficient of similarity to ordinal characters","authors":"Podani","year":"1999","venue":"Taxon","doi":null,"url":null,"internal_url":null},{"position":24,"raw":"Riley, Accessed 5 DEC 2017]","title":null,"authors":"Riley","year":"2017","venue":null,"doi":null,"url":null,"internal_url":null},{"position":25,"raw":"Jenn Riley. Understand- ing metadata. http://www.niso.org/publications/press/ UnderstandingMetadata.pdf, Accessed: 5 DEC 2017.","title":"Understanding metadata","authors":"Jenn Riley","year":"2017","venue":null,"doi":null,"url":"http://www.niso.org/publications/press/UnderstandingMetadata.pdf","internal_url":null},{"position":26,"raw":"[Shi and Malik, 2000] Jianbo Shi and Jitendra Malik. Nor- malized cuts and image segmentation. IEEE TPAMI, 22(8):888-905, 2000.","title":"Normalized cuts and image segmentation","authors":"Jianbo Shi, Jitendra Malik","year":"2000","venue":"IEEE TPAMI","doi":null,"url":null,"internal_url":null},{"position":27,"raw":"[The University of Sydney, 2018] The University of Sydney. Ashrae rp-884 adaptive model project. https://sydney.edu.au/architecture/staff/homepage/ richard de dear/ashrae rp-884.shtml, 2018.","title":"Ashrae rp-884 adaptive model project","authors":null,"year":"2018","venue":null,"doi":null,"url":"https://sydney.edu.au/architecture/staff/homepage/richarddedear/ashraerp-884.shtml","internal_url":null},{"position":28,"raw":"[Wang et al., 2009] F. Wang, X. Wang, and T. Li. Semi- supervised multi-task learning with task regularizations. In IEEE ICDM, pages 562-568, 2009. [Wikipedia, Accessed 2 DEC 2018] Wikipedia. Metadata. https://en.wikipedia.org/wiki/Metadata, Accessed 2 DEC 2018.","title":"Semisupervised multi-task learning with task regularizations","authors":"F Wang, X Wang, T Li","year":"2009","venue":"IEEE ICDM","doi":null,"url":"https://en.wikipedia.org/wiki/Metadata","internal_url":null},{"position":29,"raw":"[Xie et al., 2016] Ruobing Xie, Zhiyuan Liu, and Maosong Sun. Representation learning of knowledge graphs with hierarchical types. In IJCAI, pages 2965-2971, 2016.","title":"Representation learning of knowledge graphs with hierarchical types","authors":"Xie","year":"2016","venue":"IJCAI","doi":null,"url":null,"internal_url":"/proceedings/2016/421"},{"position":30,"raw":"[Xu and others, 2015] Jianpeng Xu et al. FORMULA: fac- torized multi-task learning for task discovery in personal- ized medical models. In SIAM, pages 496-504, 2015.","title":"FORMULA: factorized multi-task learning for task discovery in personalized medical models","authors":"Others ; Jianpeng Xu, Xu","year":"2015","venue":"SIAM","doi":null,"url":null,"internal_url":null},{"position":31,"raw":"[Zamir and others, 2018] Amir R Zamir et al. Taskonomy: Disentangling task transfer learning. In IEEE CVPR, pages 3712-3722, 2018.","title":"Taskonomy: Disentangling task transfer learning","authors":"Amir R Zamir","year":"2018","venue":"IEEE CVPR","doi":null,"url":null,"internal_url":"/proceedings/2019/871"},{"position":32,"raw":"[Zhang and Yang, 2017a] Yu Zhang and Qiang Yang. Learn- ing sparse task relations in multi-task learning. In AAAI, pages 2914-2920, 2017.","title":"Learning sparse task relations in multi-task learning","authors":"Yang Zhang, Yu Zhang, Qiang Yang","year":"2017","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":33,"raw":"[Zhang and Yang, 2017b] Yu Zhang and Qiang Yang. A survey on multi-task learning. arXiv preprint arXiv:1707.08114, 2017.","title":"A survey on multi-task learning","authors":"Yang ; Zhang, Yu Zhang, Qiang Yang","year":"2017","venue":null,"doi":null,"url":null,"internal_url":null},{"position":34,"raw":"[Zhang et al., 2016] Xianchao Zhang, Xiaotong Zhang, and Han Liu. Self-adapted multi-task clustering. In IJCAI, pages 2357-2363, 2016.","title":"Self-adapted multi-task clustering","authors":"Zhang","year":"2016","venue":"IJCAI","doi":null,"url":null,"internal_url":"/proceedings/2016/336"}],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}