Home > Research > Publications & Outputs > Synthesizing Samples fro Zero-shot Learning

Associated organisational unit

Electronic data

  • ijcai2017_submission_yuchenzeroshot

    Accepted author manuscript, 1 MB, PDF document

    Available under license: CC BY-NC: Creative Commons Attribution-NonCommercial 4.0 International License

Links

Text available via DOI:

View graph of relations

Synthesizing Samples fro Zero-shot Learning

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paperpeer-review

Published

Standard

Synthesizing Samples fro Zero-shot Learning. / Guo, Yuchen; Ding, Guiguang; Han, Jungong et al.
Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence. ed. / Carles Sierra. Melbourne: IJCAI, 2017. p. 1774-1780.

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paperpeer-review

Harvard

Guo, Y, Ding, G, Han, J & Gao, Y 2017, Synthesizing Samples fro Zero-shot Learning. in C Sierra (ed.), Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence. IJCAI, Melbourne, pp. 1774-1780, IJCAI17, 21/08/17. https://doi.org/10.24963/ijcai.2017/246

APA

Guo, Y., Ding, G., Han, J., & Gao, Y. (2017). Synthesizing Samples fro Zero-shot Learning. In C. Sierra (Ed.), Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (pp. 1774-1780). IJCAI. https://doi.org/10.24963/ijcai.2017/246

Vancouver

Guo Y, Ding G, Han J, Gao Y. Synthesizing Samples fro Zero-shot Learning. In Sierra C, editor, Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence. Melbourne: IJCAI. 2017. p. 1774-1780 doi: 10.24963/ijcai.2017/246

Author

Guo, Yuchen ; Ding, Guiguang ; Han, Jungong et al. / Synthesizing Samples fro Zero-shot Learning. Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence. editor / Carles Sierra. Melbourne : IJCAI, 2017. pp. 1774-1780

Bibtex

@inproceedings{7d2aec290c31442ead7524ddd0cb1680,
title = "Synthesizing Samples fro Zero-shot Learning",
abstract = "Zero-shot learning (ZSL) is to construct recognition models for unseen target classes that have no labeled samples for training. It utilizes the class attributes or semantic vectors as side information and transfers supervision information from related source classes with abundant labeled samples. Existing ZSL approaches adopt an intermediary embedding space to measure the similarity between a sample and the attributes of a target class to perform zero-shot classification. However, this way may suffer from the information loss caused by the embedding process and the similarity measure cannot fully make use of the data distribution. In this paper, we propose a novel approach which turns the ZSL problem into a conventional supervised learning problem by synthesizing samples for the unseen classes. Firstly, the probability distribution of an unseen class is estimated by using the knowledge from seen classes and the class attributes. Secondly, the samples are synthesized based on the distribution for the unseen class. Finally, we can train any supervised classifiers based on the synthesized samples. Extensive experiments on benchmarks demonstrate the superiority of the proposed approach to the state-of-the-art ZSL approaches.",
author = "Yuchen Guo and Guiguang Ding and Jungong Han and Yue Gao",
year = "2017",
month = aug,
day = "19",
doi = "10.24963/ijcai.2017/246",
language = "English",
pages = "1774--1780",
editor = "Carles Sierra",
booktitle = "Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence",
publisher = "IJCAI",
note = "IJCAI17 ; Conference date: 21-08-2017 Through 25-08-2017",

}

RIS

TY - GEN

T1 - Synthesizing Samples fro Zero-shot Learning

AU - Guo, Yuchen

AU - Ding, Guiguang

AU - Han, Jungong

AU - Gao, Yue

PY - 2017/8/19

Y1 - 2017/8/19

N2 - Zero-shot learning (ZSL) is to construct recognition models for unseen target classes that have no labeled samples for training. It utilizes the class attributes or semantic vectors as side information and transfers supervision information from related source classes with abundant labeled samples. Existing ZSL approaches adopt an intermediary embedding space to measure the similarity between a sample and the attributes of a target class to perform zero-shot classification. However, this way may suffer from the information loss caused by the embedding process and the similarity measure cannot fully make use of the data distribution. In this paper, we propose a novel approach which turns the ZSL problem into a conventional supervised learning problem by synthesizing samples for the unseen classes. Firstly, the probability distribution of an unseen class is estimated by using the knowledge from seen classes and the class attributes. Secondly, the samples are synthesized based on the distribution for the unseen class. Finally, we can train any supervised classifiers based on the synthesized samples. Extensive experiments on benchmarks demonstrate the superiority of the proposed approach to the state-of-the-art ZSL approaches.

AB - Zero-shot learning (ZSL) is to construct recognition models for unseen target classes that have no labeled samples for training. It utilizes the class attributes or semantic vectors as side information and transfers supervision information from related source classes with abundant labeled samples. Existing ZSL approaches adopt an intermediary embedding space to measure the similarity between a sample and the attributes of a target class to perform zero-shot classification. However, this way may suffer from the information loss caused by the embedding process and the similarity measure cannot fully make use of the data distribution. In this paper, we propose a novel approach which turns the ZSL problem into a conventional supervised learning problem by synthesizing samples for the unseen classes. Firstly, the probability distribution of an unseen class is estimated by using the knowledge from seen classes and the class attributes. Secondly, the samples are synthesized based on the distribution for the unseen class. Finally, we can train any supervised classifiers based on the synthesized samples. Extensive experiments on benchmarks demonstrate the superiority of the proposed approach to the state-of-the-art ZSL approaches.

U2 - 10.24963/ijcai.2017/246

DO - 10.24963/ijcai.2017/246

M3 - Conference contribution/Paper

SP - 1774

EP - 1780

BT - Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence

A2 - Sierra, Carles

PB - IJCAI

CY - Melbourne

T2 - IJCAI17

Y2 - 21 August 2017 through 25 August 2017

ER -