MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement

Szu-Wei Fu, Chien-Feng Liao, Yu Tsao, Shou-De Lin
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:2031-2041, 2019.

Abstract

Adversarial loss in a conditional generative adversarial network (GAN) is not designed to directly optimize evaluation metrics of a target task, and thus, may not always guide the generator in a GAN to generate data with improved metric scores. To overcome this issue, we propose a novel MetricGAN approach with an aim to optimize the generator with respect to one or multiple evaluation metrics. Moreover, based on MetricGAN, the metric scores of the generated data can also be arbitrarily specified by users. We tested the proposed MetricGAN on a speech enhancement task, which is particularly suitable to verify the proposed approach because there are multiple metrics measuring different aspects of speech signals. Moreover, these metrics are generally complex and could not be fully optimized by Lp or conventional adversarial losses.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-fu19b, title = {{M}etric{GAN}: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement}, author = {Fu, Szu-Wei and Liao, Chien-Feng and Tsao, Yu and Lin, Shou-De}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {2031--2041}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/fu19b/fu19b.pdf}, url = {https://proceedings.mlr.press/v97/fu19b.html}, abstract = {Adversarial loss in a conditional generative adversarial network (GAN) is not designed to directly optimize evaluation metrics of a target task, and thus, may not always guide the generator in a GAN to generate data with improved metric scores. To overcome this issue, we propose a novel MetricGAN approach with an aim to optimize the generator with respect to one or multiple evaluation metrics. Moreover, based on MetricGAN, the metric scores of the generated data can also be arbitrarily specified by users. We tested the proposed MetricGAN on a speech enhancement task, which is particularly suitable to verify the proposed approach because there are multiple metrics measuring different aspects of speech signals. Moreover, these metrics are generally complex and could not be fully optimized by Lp or conventional adversarial losses.} }
Endnote
%0 Conference Paper %T MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement %A Szu-Wei Fu %A Chien-Feng Liao %A Yu Tsao %A Shou-De Lin %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-fu19b %I PMLR %P 2031--2041 %U https://proceedings.mlr.press/v97/fu19b.html %V 97 %X Adversarial loss in a conditional generative adversarial network (GAN) is not designed to directly optimize evaluation metrics of a target task, and thus, may not always guide the generator in a GAN to generate data with improved metric scores. To overcome this issue, we propose a novel MetricGAN approach with an aim to optimize the generator with respect to one or multiple evaluation metrics. Moreover, based on MetricGAN, the metric scores of the generated data can also be arbitrarily specified by users. We tested the proposed MetricGAN on a speech enhancement task, which is particularly suitable to verify the proposed approach because there are multiple metrics measuring different aspects of speech signals. Moreover, these metrics are generally complex and could not be fully optimized by Lp or conventional adversarial losses.
APA
Fu, S., Liao, C., Tsao, Y. & Lin, S.. (2019). MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:2031-2041 Available from https://proceedings.mlr.press/v97/fu19b.html.

Related Material