Differentially Private Policy Evaluation

Borja Balle, Maziar Gomrokchi, Doina Precup
Proceedings of The 33rd International Conference on Machine Learning, PMLR 48:2130-2138, 2016.

Abstract

We present the first differentially private algorithms for reinforcement learning, which apply to the task of evaluating a fixed policy. We establish two approaches for achieving differential privacy, provide a theoretical analysis of the privacy and utility of the two algorithms, and show promising results on simple empirical examples.

Cite this Paper


BibTeX
@InProceedings{pmlr-v48-balle16, title = {Differentially Private Policy Evaluation}, author = {Balle, Borja and Gomrokchi, Maziar and Precup, Doina}, booktitle = {Proceedings of The 33rd International Conference on Machine Learning}, pages = {2130--2138}, year = {2016}, editor = {Balcan, Maria Florina and Weinberger, Kilian Q.}, volume = {48}, series = {Proceedings of Machine Learning Research}, address = {New York, New York, USA}, month = {20--22 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v48/balle16.pdf}, url = { http://proceedings.mlr.press/v48/balle16.html }, abstract = {We present the first differentially private algorithms for reinforcement learning, which apply to the task of evaluating a fixed policy. We establish two approaches for achieving differential privacy, provide a theoretical analysis of the privacy and utility of the two algorithms, and show promising results on simple empirical examples.} }
Endnote
%0 Conference Paper %T Differentially Private Policy Evaluation %A Borja Balle %A Maziar Gomrokchi %A Doina Precup %B Proceedings of The 33rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Maria Florina Balcan %E Kilian Q. Weinberger %F pmlr-v48-balle16 %I PMLR %P 2130--2138 %U http://proceedings.mlr.press/v48/balle16.html %V 48 %X We present the first differentially private algorithms for reinforcement learning, which apply to the task of evaluating a fixed policy. We establish two approaches for achieving differential privacy, provide a theoretical analysis of the privacy and utility of the two algorithms, and show promising results on simple empirical examples.
RIS
TY - CPAPER TI - Differentially Private Policy Evaluation AU - Borja Balle AU - Maziar Gomrokchi AU - Doina Precup BT - Proceedings of The 33rd International Conference on Machine Learning DA - 2016/06/11 ED - Maria Florina Balcan ED - Kilian Q. Weinberger ID - pmlr-v48-balle16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 48 SP - 2130 EP - 2138 L1 - http://proceedings.mlr.press/v48/balle16.pdf UR - http://proceedings.mlr.press/v48/balle16.html AB - We present the first differentially private algorithms for reinforcement learning, which apply to the task of evaluating a fixed policy. We establish two approaches for achieving differential privacy, provide a theoretical analysis of the privacy and utility of the two algorithms, and show promising results on simple empirical examples. ER -
APA
Balle, B., Gomrokchi, M. & Precup, D.. (2016). Differentially Private Policy Evaluation. Proceedings of The 33rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 48:2130-2138 Available from http://proceedings.mlr.press/v48/balle16.html .

Related Material