3,935 Hits in 4.5 sec

Local Differential Privacy for Regret Minimization in Reinforcement Learning [article]

Evrard Garcelon, Vianney Perchet, Ciara Pike-Burke, Matteo Pirotta
2021 arXiv   pre-print
We formulate this notion of privacy for RL by leveraging the local differential privacy (LDP) framework.  ...  We establish a lower bound for regret minimization in finite-horizon MDPs with LDP guarantees which shows that guaranteeing privacy has a multiplicative effect on the regret.  ...  G Posterior Sampling for Local Differential Privacy The Posterior Sampling for Reinforcement Learning algorithm [PSRL, Osband et al., 2013 ] is a Thompson Sampling based algorithm for Reinforcement Learning  ... 
arXiv:2010.07778v3 fatcat:76bncyh47zgr3bvtu5qf52gxnm

Differentially Private Regret Minimization in Episodic Markov Decision Processes [article]

Sayak Ray Chowdhury, Xingyu Zhou
2021 arXiv   pre-print
We study regret minimization in finite horizon tabular Markov decision processes (MDPs) under the constraints of differential privacy (DP).  ...  This is motivated by the widespread applications of reinforcement learning (RL) in real-world sequential decision making problems, where protecting users' sensitive and private information is becoming  ...  Local differentially private regret minimization in reinforcement learning. arXiv preprint arXiv:2010.07778, 2020. Cynthia Dwork. Differential privacy: A survey of results.  ... 
arXiv:2112.10599v1 fatcat:nuiwxweo7vcajhfbjeiilo5zle

Shuffle Private Linear Contextual Bandits [article]

Sayak Ray Chowdhury, Xingyu Zhou
2022 arXiv   pre-print
However, there remains a fundamental gap in the utility achieved by learning algorithms under these two privacy models, e.g., Õ(√(T)) regret in the central model as compared to Õ(T^3/4) regret in the local  ...  Differential privacy (DP) has been recently introduced to linear contextual bandits to formally address the privacy concerns in its associated personalized services to participating users (e.g., recommendations  ...  Locally differentially private (contextual) bandits learning. In NeurIPS, 2020.  ... 
arXiv:2202.05567v1 fatcat:z5mnuy4jtvhubfitvvfccfpvke

Differentially Private Reinforcement Learning with Linear Function Approximation

Xingyu Zhou
2022 Proceedings of the ACM on Measurement and Analysis of Computing Systems  
Motivated by the wide adoption of reinforcement learning (RL) in real-world personalized services, where users' sensitive and private information needs to be protected, we study regret minimization in  ...  a building block for general private reinforcement learning.  ...  in reinforcement learning.  ... 
doi:10.1145/3508028 fatcat:v4u3idqfwvcd5megi3gdroalli

Locally Differentially Private Reinforcement Learning for Linear Mixture Markov Decision Processes [article]

Chonghua Liao and Jiafan He and Quanquan Gu
2021 arXiv   pre-print
To protect the users' privacy, privacy-preserving RL algorithms are in demand. In this paper, we study RL with linear function approximation and local differential privacy (LDP) guarantees.  ...  We propose a novel (ε, δ)-LDP algorithm for learning a class of Markov decision processes (MDPs) dubbed linear mixture MDPs, and obtains an 𝒪̃( d^5/4H^7/4T^3/4(log(1/δ))^1/4√(1/ε)) regret, where d is  ...  (Local) Differential Privacy In this subsection, we introduce the standard definition of differential privacy (Dwork et al., 2006) and local differential privacy (Kasiviswanathan et al., 2011; Duchi  ... 
arXiv:2110.10133v1 fatcat:gvvkdlzxr5eylcawjdp5xrmgcy

Differentially Private Meta-Learning [article]

Jeffrey Li, Mikhail Khodak, Sebastian Caldas, Ameet Talwalkar
2020 arXiv   pre-print
Parameter-transfer is a well-known and versatile approach for meta-learning, with applications including few-shot learning, federated learning, and reinforcement learning.  ...  We then propose a new differentially private algorithm for gradient-based parameter transfer that not only satisfies this privacy requirement but also retains provable transfer learning guarantees in convex  ...  ACKNOWLEDGMENTS This work was supported in part by DARPA FA875017C0141, the National Science Foundation grants IIS1618714, IIS1705121, and IIS1838017, an Okawa Grant, a Google Faculty Award, an Amazon  ... 
arXiv:1909.05830v2 fatcat:e37kmf6fsrdutcn2w2p5qarvxu

Multi-Armed Bandits with Local Differential Privacy [article]

Wenbo Ren, Xingyu Zhou, Jia Liu, Ness B. Shroff
2020 arXiv   pre-print
This paper investigates the problem of regret minimization for multi-armed bandit (MAB) problems with local differential privacy (LDP) guarantee.  ...  To handle this dilemma, we adopt differential privacy and study the regret upper and lower bounds for MAB algorithms with a given LDP guarantee.  ...  Mathematically, the (pseudo) regret is defined as R(T ) := T µ * − T t=1 E[µ A t ] = E[ a∈[n] N T a ∆ a ]. The goal of the agent is to minimize the regret. Local differential privacy.  ... 
arXiv:2007.03121v1 fatcat:eehdco6p4ndpffatetx47o4nzi

Private Reinforcement Learning with PAC and Regret Guarantees [article]

Giuseppe Vietri, Borja Balle, Akshay Krishnamurthy, Zhiwei Steven Wu
2020 arXiv   pre-print
Finally, we present lower bounds on sample complexity and regret for reinforcement learning subject to JDP.  ...  Our algorithm only pays for a moderate privacy cost on exploration: in comparison to the non-private bounds, the privacy parameter only appears in lower-order terms.  ...  In this paper we contribute to the study of reinforcement learning from the lens of differential privacy.  ... 
arXiv:2009.09052v1 fatcat:piv3gjndg5g5lfnmnfo5qsxfuq

Corrupt Bandits for Preserving Local Privacy [article]

Pratik Gajane, Tanguy Urvoy, Emilie Kaufmann
2017 arXiv   pre-print
We also provide the appropriate corruption parameters to guarantee a desired level of local privacy and analyze how this impacts the regret.  ...  We provide a lower bound on the expected regret of any bandit algorithm in this corrupted setting.  ...  In this article, we consider local differential privacy.  ... 
arXiv:1708.05033v2 fatcat:wpvawzedend5tpcgft272z7v3q

Local Differential Privacy and Its Applications: A Comprehensive Survey [article]

Mengmeng Yang, Lingjuan Lyu, Jun Zhao, Tianqing Zhu, Kwok-Yan Lam
2020 arXiv   pre-print
Local differential privacy (LDP), as a strong privacy tool, has been widely deployed in the real world in recent years.  ...  We discuss the practical deployment of local differential privacy and explore its application in various domains.  ...  [91] propose a local differential privacy algorithm based on asynchronous advantage actor-critic (A3C) for distributed reinforcement learning to obtain a robust policy.  ... 
arXiv:2008.03686v1 fatcat:l7z3gip2ivdmvin7lraxd4vciy

Byzantine-Robust Federated Linear Bandits [article]

Ali Jadbabaie, Haochuan Li, Jian Qian, Yi Tian
2022 arXiv   pre-print
In this paper, we study a linear bandit optimization problem in a federated setting where a large collection of distributed agents collaboratively learn a common linear bandit model.  ...  Finally, if the level of corruption is known to be small, we show that using the geometric median of mean oracle for robust aggregation further improves the regret bound.  ...  Differential privacy. Privacy issues are also important for distributed systems.  ... 
arXiv:2204.01155v1 fatcat:64deqod4xfcmnbj34jtybxvzpa

Federated Bandit: A Gossiping Approach [article]

Zhaowei Zhu, Jingxuan Zhu, Ji Liu, Yang Liu
2020 arXiv   pre-print
We then propose Fed_UCB, a differentially private version of Gossip_UCB, in which the agents preserve ϵ-differential privacy of their local data while achieving O(max{poly(N,M)/ϵlog^2.5 T, poly(N,M) (log_λ  ...  We show that Gossip_UCB successfully adapts local bandit learning into a global gossiping process for sharing information among connected agents, and achieves guaranteed regret at the order of O(max{poly  ...  through gossiping, and its differentially private variant, Fed_UCB, for preserving -differential privacy of the agents' local data.  ... 
arXiv:2010.12763v1 fatcat:skq7homjy5cs5phj3ikbcda2r4

Trusted AI in Multi-agent Systems: An Overview of Privacy and Security for Distributed Learning [article]

Chuan Ma, Jun Li, Kang Wei, Bo Liu, Ming Ding, Long Yuan, Zhu Han, H. Vincent Poor
2022 arXiv   pre-print
Finally, we complete the survey by providing an outlook on the challenges and possible directions for future research in this critical area.  ...  In this paper, we provide a survey of the emerging security and privacy risks of distributed ML from a unique perspective of information exchange levels, which are defined according to the key steps of  ...  Differential Privacy Differential privacy (DP) is a standard definition for privacy estimation [130] .  ... 
arXiv:2202.09027v2 fatcat:hlu7bopcjrc6zjn2pct57utufy

A Comprehensive Survey of Incentive Mechanism for Federated Learning [article]

Rongfei Zeng, Chao Zeng, Xingwei Wang, Bo Li, Xiaowen Chu
2021 arXiv   pre-print
In this paper, we present a comprehensive survey of incentive schemes for federate learning.  ...  Specifically, we identify the incentive problem in federated learning and then provide a taxonomy for various schemes.  ...  In sum, reinforcement learning can be innovatively applied for the incentive design in FL.  ... 
arXiv:2106.15406v1 fatcat:7v7adaw5wnbldhykma6mhgpnty

Interactive Sensing and Decision Making in Social Networks

Vikram Krishnamurthy
2014 Foundations and Trends® in Signal Processing  
This monograph provides a survey, tutorial development, and discussion of four highly stylized examples: social learning for interactive sensing; tracking the degree distribution of social networks; sensing  ...  and information diffusion; and coordination of decision making via game-theoretic learning.  ...  In social learning, each agent chooses its action by optimizing its local utility function.  ... 
doi:10.1561/2000000048 fatcat:5hr4ebohczhrlolw4whenlicj4
« Previous Showing results 1 — 15 out of 3,935 results