Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs

Michael Gygli, Mohammad Norouzi, Anelia Angelova
Proceedings of the 34th International Conference on Machine Learning, PMLR 70:1341-1351, 2017.

Abstract

We approach structured output prediction by optimizing a deep value network (DVN) to precisely estimate the task loss on different output configurations for a given input. Once the model is trained, we perform inference by gradient descent on the continuous relaxations of the output variables to find outputs with promising scores from the value network. When applied to image segmentation, the value network takes an image and a segmentation mask as inputs and predicts a scalar estimating the intersection over union between the input and ground truth masks. For multi-label classification, the DVN’s objective is to correctly predict the F1 score for any potential label configuration. The DVN framework achieves the state-of-the-art results on multi-label prediction and image segmentation benchmarks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v70-gygli17a, title = {Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs}, author = {Michael Gygli and Mohammad Norouzi and Anelia Angelova}, booktitle = {Proceedings of the 34th International Conference on Machine Learning}, pages = {1341--1351}, year = {2017}, editor = {Precup, Doina and Teh, Yee Whye}, volume = {70}, series = {Proceedings of Machine Learning Research}, month = {06--11 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v70/gygli17a/gygli17a.pdf}, url = {https://proceedings.mlr.press/v70/gygli17a.html}, abstract = {We approach structured output prediction by optimizing a deep value network (DVN) to precisely estimate the task loss on different output configurations for a given input. Once the model is trained, we perform inference by gradient descent on the continuous relaxations of the output variables to find outputs with promising scores from the value network. When applied to image segmentation, the value network takes an image and a segmentation mask as inputs and predicts a scalar estimating the intersection over union between the input and ground truth masks. For multi-label classification, the DVN’s objective is to correctly predict the F1 score for any potential label configuration. The DVN framework achieves the state-of-the-art results on multi-label prediction and image segmentation benchmarks.} }
Endnote
%0 Conference Paper %T Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs %A Michael Gygli %A Mohammad Norouzi %A Anelia Angelova %B Proceedings of the 34th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2017 %E Doina Precup %E Yee Whye Teh %F pmlr-v70-gygli17a %I PMLR %P 1341--1351 %U https://proceedings.mlr.press/v70/gygli17a.html %V 70 %X We approach structured output prediction by optimizing a deep value network (DVN) to precisely estimate the task loss on different output configurations for a given input. Once the model is trained, we perform inference by gradient descent on the continuous relaxations of the output variables to find outputs with promising scores from the value network. When applied to image segmentation, the value network takes an image and a segmentation mask as inputs and predicts a scalar estimating the intersection over union between the input and ground truth masks. For multi-label classification, the DVN’s objective is to correctly predict the F1 score for any potential label configuration. The DVN framework achieves the state-of-the-art results on multi-label prediction and image segmentation benchmarks.
APA
Gygli, M., Norouzi, M. & Angelova, A.. (2017). Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs. Proceedings of the 34th International Conference on Machine Learning, in Proceedings of Machine Learning Research 70:1341-1351 Available from https://proceedings.mlr.press/v70/gygli17a.html.

Related Material