Data Valuation using Reinforcement Learning

Jinsung Yoon, Sercan Arik, Tomas Pfister
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:10842-10851, 2020.

Abstract

Quantifying the value of data is a fundamental problem in machine learning and has multiple important use cases: (1) building insights about the dataset and task, (2) domain adaptation, (3) corrupted sample discovery, and (4) robust learning. We propose Data Valuation using Reinforcement Learning (DVRL), to adaptively learn data values jointly with the predictor model. DVRL uses a data value estimator (DVE) to learn how likely each datum is used in training of the predictor model. DVE is trained using a reinforcement signal that reflects performance on the target task. We demonstrate that DVRL yields superior data value estimates compared to alternative methods across numerous datasets and application scenarios. The corrupted sample discovery performance of DVRL is close to optimal in many regimes (i.e. as if the noisy samples were known apriori), and for domain adaptation and robust learning DVRL significantly outperforms state-of-the-art by 14.6% and 10.8%, respectively.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-yoon20a, title = {Data Valuation using Reinforcement Learning}, author = {Yoon, Jinsung and Arik, Sercan and Pfister, Tomas}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {10842--10851}, year = {2020}, editor = {Hal Daumé III and Aarti Singh}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/yoon20a/yoon20a.pdf}, url = { http://proceedings.mlr.press/v119/yoon20a.html }, abstract = {Quantifying the value of data is a fundamental problem in machine learning and has multiple important use cases: (1) building insights about the dataset and task, (2) domain adaptation, (3) corrupted sample discovery, and (4) robust learning. We propose Data Valuation using Reinforcement Learning (DVRL), to adaptively learn data values jointly with the predictor model. DVRL uses a data value estimator (DVE) to learn how likely each datum is used in training of the predictor model. DVE is trained using a reinforcement signal that reflects performance on the target task. We demonstrate that DVRL yields superior data value estimates compared to alternative methods across numerous datasets and application scenarios. The corrupted sample discovery performance of DVRL is close to optimal in many regimes (i.e. as if the noisy samples were known apriori), and for domain adaptation and robust learning DVRL significantly outperforms state-of-the-art by 14.6% and 10.8%, respectively.} }
Endnote
%0 Conference Paper %T Data Valuation using Reinforcement Learning %A Jinsung Yoon %A Sercan Arik %A Tomas Pfister %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-yoon20a %I PMLR %P 10842--10851 %U http://proceedings.mlr.press/v119/yoon20a.html %V 119 %X Quantifying the value of data is a fundamental problem in machine learning and has multiple important use cases: (1) building insights about the dataset and task, (2) domain adaptation, (3) corrupted sample discovery, and (4) robust learning. We propose Data Valuation using Reinforcement Learning (DVRL), to adaptively learn data values jointly with the predictor model. DVRL uses a data value estimator (DVE) to learn how likely each datum is used in training of the predictor model. DVE is trained using a reinforcement signal that reflects performance on the target task. We demonstrate that DVRL yields superior data value estimates compared to alternative methods across numerous datasets and application scenarios. The corrupted sample discovery performance of DVRL is close to optimal in many regimes (i.e. as if the noisy samples were known apriori), and for domain adaptation and robust learning DVRL significantly outperforms state-of-the-art by 14.6% and 10.8%, respectively.
APA
Yoon, J., Arik, S. & Pfister, T.. (2020). Data Valuation using Reinforcement Learning. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:10842-10851 Available from http://proceedings.mlr.press/v119/yoon20a.html .

Related Material