抄録
In cloud computing, methods to perform machine learning while maintaining data security have been widely studied. One of such methods is the secret computation method, which achieves high confidentiality by decomposing data and parameters for machine learning and performing distributed processing. The author has proposed an algorithm that realizes the BP method, one of the machine learning methods, by allowing each terminal in a cloud system to act autonomously without a central terminal. However, the number of parameters of this method increases as the number of data partitions increases. In this paper, we propose a method in which the number of parameters does not increase from the viewpoint of machine learning, even if the number of data partitions increases, by devising an update formula for the parameters in each terminal. Numerical experiments are conducted to demonstrate the effectiveness of the proposed method．
