We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Analog Q-learning Methods for Secure Multiparty Computation.
- Authors
Hirofumi Miyajima; Noritaka Shigei; Hiromi Miyajima; Norio Shiratori
- Abstract
One problem in cloud computing system is how to conceal individual information. Although encryption technology is one of methods to solve the problem, the computation time required for encryption and decryption as the amount of data increases is a bottleneck. On the other hand, the secret processing of SMC, by reducing the amount of data processed by partitioning data, achieves confidentiality and speeding up. Compared with encryption technology, SMC can realize highspeed and secret processing, but in order to perform it many servers are required. Therefore, an easily method using SMC and simple encryption has been proposed. Several algorithms related to supervised, unsupervised and reinforcement learning have been proposed so far as the methods of SMC on machine learning. Since there is no learning data in reinforcement learning, the result is obtained on the client without informing the solution system (parameter) to any server. Algorithms for Q learning and PS learning in digital model have been proposed so far but no results on analog model have been obtained yet. In this paper, an algorithm of Q learning in analog model is proposed. Moreover, the effectiveness of the result is shown by numerical simulation. of this column.
- Subjects
CLOUD computing; ARTIFICIAL intelligence; COMPUTER algorithms; X-ray diffraction; COMPUTER security
- Publication
IAENG International Journal of Computer Science, 2018, Vol 45, Issue 4, p71
- ISSN
1819-656X
- Publication type
Article