Autoencoding beyond pixels using a learned similarity metric
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Standard
Autoencoding beyond pixels using a learned similarity metric. / Larsen, Anders Boesen Lindbo; Sønderby, Søren Kaae; Larochelle, Hugo; Winther, Ole.
Proceedings of The 33rd International Conference on Machine Learning. ed. / Maria Florina Balcan; Kilian Q. Weinberger. 2016. p. 1558–1566 (JMLR: Workshop and Conference Proceedings, Vol. 48).Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Autoencoding beyond pixels using a learned similarity metric
AU - Larsen, Anders Boesen Lindbo
AU - Sønderby, Søren Kaae
AU - Larochelle, Hugo
AU - Winther, Ole
N1 - Conference code: 33
PY - 2016
Y1 - 2016
N2 - We present an autoencoder that leverages learned representations to better measure similarities in data space. By combining a variational autoencoder (VAE) with a generative adversarial network (GAN) we can use learned feature representations in the GAN discriminator as basis for the VAE reconstruction objective. Thereby, we replace element-wise errors with feature-wise errors to better capture the data distribution while offering invariance towards e.g. translation. We apply our method to images of faces and show that it outperforms VAEs with element-wise similarity measures in terms of visual fidelity. Moreover, we show that the method learns an embedding in which high-level abstract visual features (e.g. wearing glasses) can be modified using simple arithmetic.
AB - We present an autoencoder that leverages learned representations to better measure similarities in data space. By combining a variational autoencoder (VAE) with a generative adversarial network (GAN) we can use learned feature representations in the GAN discriminator as basis for the VAE reconstruction objective. Thereby, we replace element-wise errors with feature-wise errors to better capture the data distribution while offering invariance towards e.g. translation. We apply our method to images of faces and show that it outperforms VAEs with element-wise similarity measures in terms of visual fidelity. Moreover, we show that the method learns an embedding in which high-level abstract visual features (e.g. wearing glasses) can be modified using simple arithmetic.
M3 - Article in proceedings
AN - SCOPUS:84999041243
T3 - JMLR: Workshop and Conference Proceedings
SP - 1558
EP - 1566
BT - Proceedings of The 33rd International Conference on Machine Learning
A2 - Balcan, Maria Florina
A2 - Weinberger, Kilian Q.
Y2 - 19 June 2016 through 24 June 2016
ER -
ID: 171660204