An Embarrassingly Simple Approach to Zero-Shot Learning. Romera-Paredes, B. & Torr, P. H. S. In Feris, R. S., Lampert, C., & Parikh, D., editors, Visual Attributes, pages 11–30. Springer International Publishing, Cham, 2017. ZSCC: NoCitationData[s0] Series Title: Advances in Computer Vision and Pattern Recognition
Paper doi abstract bibtex Zero-shot learning consists in learning how to recognise new concepts by just having a description of them. Many sophisticated approaches have been proposed to address the challenges this problem comprises. In this paper we describe a zero-shot learning approach that can be implemented in just one line of code, yet it is able to outperform state of the art approaches on standard datasets. The approach is based on a more general framework which models the relationships between features, attributes, and classes as a two linear layers network, where the weights of the top layer are not learned but are given by the environment. We further provide a learning bound on the generalisation error of this kind of approaches, by casting them as domain adaptation methods. In experiments carried out on three standard real datasets, we found that our approach is able to perform significantly better than the state of art on all of them, obtaining a ratio of improvement up to 17%.
@incollection{feris_embarrassingly_2017,
address = {Cham},
title = {An {Embarrassingly} {Simple} {Approach} to {Zero}-{Shot} {Learning}},
isbn = {978-3-319-50075-1 978-3-319-50077-5},
url = {http://link.springer.com/10.1007/978-3-319-50077-5_2},
abstract = {Zero-shot learning consists in learning how to recognise new concepts by just having a description of them. Many sophisticated approaches have been proposed to address the challenges this problem comprises. In this paper we describe a zero-shot learning approach that can be implemented in just one line of code, yet it is able to outperform state of the art approaches on standard datasets. The approach is based on a more general framework which models the relationships between features, attributes, and classes as a two linear layers network, where the weights of the top layer are not learned but are given by the environment. We further provide a learning bound on the generalisation error of this kind of approaches, by casting them as domain adaptation methods. In experiments carried out on three standard real datasets, we found that our approach is able to perform significantly better than the state of art on all of them, obtaining a ratio of improvement up to 17\%.},
language = {en},
urldate = {2021-01-16},
booktitle = {Visual {Attributes}},
publisher = {Springer International Publishing},
author = {Romera-Paredes, Bernardino and Torr, Philip H. S.},
editor = {Feris, Rogerio Schmidt and Lampert, Christoph and Parikh, Devi},
year = {2017},
doi = {10.1007/978-3-319-50077-5_2},
note = {ZSCC: NoCitationData[s0]
Series Title: Advances in Computer Vision and Pattern Recognition},
keywords = {💡打电话},
pages = {11--30},
}
Downloads: 0
{"_id":"7GQicvQSr9CqeyaB6","bibbaseid":"romeraparedes-torr-anembarrassinglysimpleapproachtozeroshotlearning-2017","author_short":["Romera-Paredes, B.","Torr, P. H. S."],"bibdata":{"bibtype":"incollection","type":"incollection","address":"Cham","title":"An Embarrassingly Simple Approach to Zero-Shot Learning","isbn":"978-3-319-50075-1 978-3-319-50077-5","url":"http://link.springer.com/10.1007/978-3-319-50077-5_2","abstract":"Zero-shot learning consists in learning how to recognise new concepts by just having a description of them. Many sophisticated approaches have been proposed to address the challenges this problem comprises. In this paper we describe a zero-shot learning approach that can be implemented in just one line of code, yet it is able to outperform state of the art approaches on standard datasets. The approach is based on a more general framework which models the relationships between features, attributes, and classes as a two linear layers network, where the weights of the top layer are not learned but are given by the environment. We further provide a learning bound on the generalisation error of this kind of approaches, by casting them as domain adaptation methods. In experiments carried out on three standard real datasets, we found that our approach is able to perform significantly better than the state of art on all of them, obtaining a ratio of improvement up to 17%.","language":"en","urldate":"2021-01-16","booktitle":"Visual Attributes","publisher":"Springer International Publishing","author":[{"propositions":[],"lastnames":["Romera-Paredes"],"firstnames":["Bernardino"],"suffixes":[]},{"propositions":[],"lastnames":["Torr"],"firstnames":["Philip","H.","S."],"suffixes":[]}],"editor":[{"propositions":[],"lastnames":["Feris"],"firstnames":["Rogerio","Schmidt"],"suffixes":[]},{"propositions":[],"lastnames":["Lampert"],"firstnames":["Christoph"],"suffixes":[]},{"propositions":[],"lastnames":["Parikh"],"firstnames":["Devi"],"suffixes":[]}],"year":"2017","doi":"10.1007/978-3-319-50077-5_2","note":"ZSCC: NoCitationData[s0] Series Title: Advances in Computer Vision and Pattern Recognition","keywords":"💡打电话","pages":"11–30","bibtex":"@incollection{feris_embarrassingly_2017,\n\taddress = {Cham},\n\ttitle = {An {Embarrassingly} {Simple} {Approach} to {Zero}-{Shot} {Learning}},\n\tisbn = {978-3-319-50075-1 978-3-319-50077-5},\n\turl = {http://link.springer.com/10.1007/978-3-319-50077-5_2},\n\tabstract = {Zero-shot learning consists in learning how to recognise new concepts by just having a description of them. Many sophisticated approaches have been proposed to address the challenges this problem comprises. In this paper we describe a zero-shot learning approach that can be implemented in just one line of code, yet it is able to outperform state of the art approaches on standard datasets. The approach is based on a more general framework which models the relationships between features, attributes, and classes as a two linear layers network, where the weights of the top layer are not learned but are given by the environment. We further provide a learning bound on the generalisation error of this kind of approaches, by casting them as domain adaptation methods. In experiments carried out on three standard real datasets, we found that our approach is able to perform significantly better than the state of art on all of them, obtaining a ratio of improvement up to 17\\%.},\n\tlanguage = {en},\n\turldate = {2021-01-16},\n\tbooktitle = {Visual {Attributes}},\n\tpublisher = {Springer International Publishing},\n\tauthor = {Romera-Paredes, Bernardino and Torr, Philip H. S.},\n\teditor = {Feris, Rogerio Schmidt and Lampert, Christoph and Parikh, Devi},\n\tyear = {2017},\n\tdoi = {10.1007/978-3-319-50077-5_2},\n\tnote = {ZSCC: NoCitationData[s0] \nSeries Title: Advances in Computer Vision and Pattern Recognition},\n\tkeywords = {💡打电话},\n\tpages = {11--30},\n}\n\n","author_short":["Romera-Paredes, B.","Torr, P. H. S."],"editor_short":["Feris, R. S.","Lampert, C.","Parikh, D."],"key":"feris_embarrassingly_2017","id":"feris_embarrassingly_2017","bibbaseid":"romeraparedes-torr-anembarrassinglysimpleapproachtozeroshotlearning-2017","role":"author","urls":{"Paper":"http://link.springer.com/10.1007/978-3-319-50077-5_2"},"keyword":["💡打电话"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"incollection","biburl":"https://bibbase.org/zotero/wybert","dataSources":["TJkbwzD8s2wCxBy6Y"],"keywords":["💡打电话"],"search_terms":["embarrassingly","simple","approach","zero","shot","learning","romera-paredes","torr"],"title":"An Embarrassingly Simple Approach to Zero-Shot Learning","year":2017}