An embarrassingly simple approach to zero-shot learning
Zero-shot learning consists in learning how to recognize 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 l...
| Main Authors: | , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
PMLR
2015
|
| _version_ | 1811141273012142080 |
|---|---|
| author | Romera-Paredes, B Torr, PHS |
| author_facet | Romera-Paredes, B Torr, PHS |
| author_sort | Romera-Paredes, B |
| collection | OXFORD |
| description | Zero-shot learning consists in learning how to recognize 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 generalization 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%. |
| first_indexed | 2024-09-25T04:35:15Z |
| format | Conference item |
| id | oxford-uuid:8e5fe03b-62d6-49a1-847f-0dc319b16b49 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-09-25T04:35:15Z |
| publishDate | 2015 |
| publisher | PMLR |
| record_format | dspace |
| spelling | oxford-uuid:8e5fe03b-62d6-49a1-847f-0dc319b16b492024-09-12T16:23:36ZAn embarrassingly simple approach to zero-shot learningConference itemhttp://purl.org/coar/resource_type/c_5794uuid:8e5fe03b-62d6-49a1-847f-0dc319b16b49EnglishSymplectic ElementsPMLR2015Romera-Paredes, BTorr, PHSZero-shot learning consists in learning how to recognize 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 generalization 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%. |
| spellingShingle | Romera-Paredes, B Torr, PHS An embarrassingly simple approach to zero-shot learning |
| title | An embarrassingly simple approach to zero-shot learning |
| title_full | An embarrassingly simple approach to zero-shot learning |
| title_fullStr | An embarrassingly simple approach to zero-shot learning |
| title_full_unstemmed | An embarrassingly simple approach to zero-shot learning |
| title_short | An embarrassingly simple approach to zero-shot learning |
| title_sort | embarrassingly simple approach to zero shot learning |
| work_keys_str_mv | AT romeraparedesb anembarrassinglysimpleapproachtozeroshotlearning AT torrphs anembarrassinglysimpleapproachtozeroshotlearning AT romeraparedesb embarrassinglysimpleapproachtozeroshotlearning AT torrphs embarrassinglysimpleapproachtozeroshotlearning |