文章基本信息

标题：Privacy Preserving Q-learning in the Analog Model for Secure Multiparty Computation
本地全文：下载
作者：Hirofumi Miyajima ; Noritaka Shigei ; Hiromi Miyajima 等
期刊名称：Lecture Notes in Engineering and Computer Science
印刷版ISSN：2078-0958
电子版ISSN：2078-0966
出版年度：2018
卷号：2233&2234
页码：374-379
出版社：Newswood and International Association of Engineers
摘要：Many studies have been done with the security of cloud computing. Though data encryption is a typical approach, high computing complexity for encryption and decryption of data is needed. Therefore, safe system for distributed processing with secure data attracts attention, and a lot of studies have been done with them. SMC (Secure Multiparty Computation) is one of these methods. So far, most of works for ML (Machine Learning) with SMC are ones with supervised and unsupervised learning such as BP (Back Propagation) and K-means methods. Then, in the case where learning data does not exist explicitly like reinforcement learning (RL), how should it be done? We already proposed some algorithms of Q-learning and PS (Profit Sharing) learning for SMC in previous papers. However, they were methods using digital models. On the other hand, solutions for analog models are desired in the real world. In this paper, we propose SMC algorithms for Q-learning in the analog model and show their effectiveness. The idea is that in the digital model, only one behavior is selected at each time, whereas in the analog model it is decided as a combination of a plural of weighted actions.
关键词：cloud computing; secure multiparty computation; Q;learning; analog model;