An embarrassingly simple approach to zero-shot learning

Bernardino Romera-Paredes, Philip Torr
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:2152-2161, 2015.

Abstract

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%.

Cite this Paper

BibTeX
@InProceedings{pmlr-v37-romera-paredes15, title = {An embarrassingly simple approach to zero-shot learning}, author = {Romera-Paredes, Bernardino and Torr, Philip}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {2152--2161}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/romera-paredes15.pdf}, url = {https://proceedings.mlr.press/v37/romera-paredes15.html}, abstract = {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%.} }
Endnote
%0 Conference Paper %T An embarrassingly simple approach to zero-shot learning %A Bernardino Romera-Paredes %A Philip Torr %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-romera-paredes15 %I PMLR %P 2152--2161 %U https://proceedings.mlr.press/v37/romera-paredes15.html %V 37 %X 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%.
RIS
TY - CPAPER TI - An embarrassingly simple approach to zero-shot learning AU - Bernardino Romera-Paredes AU - Philip Torr BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-romera-paredes15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 2152 EP - 2161 L1 - http://proceedings.mlr.press/v37/romera-paredes15.pdf UR - https://proceedings.mlr.press/v37/romera-paredes15.html AB - 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%. ER -
APA
Romera-Paredes, B. & Torr, P.. (2015). An embarrassingly simple approach to zero-shot learning. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:2152-2161 Available from https://proceedings.mlr.press/v37/romera-paredes15.html.

Related Material