Learning algorithms for vector quantization using vertically partitioned data with IoT

Abstract

The use of cloud computing systems that support AI is expanding. On the other hand, as the number of terminals and devices connected to the system increases, the decline in capacity is becoming apparent. Edge (or fog) computing system is known as one for improving a conventional cloud system. Then, how should machine learning be realized on the edge system? SMC (Secure Multiparty Computation) is known as one model to perform secure learning methods. Horizontally and vertically partitioned data are used as data structure for SMC. There have been proposed some methods of machine learning using horizontally partitioned data of SMC on the edge system. On the other hand, few studies have been done on methods using the vertically partitioned data. In this paper, fast and secure vector quantization algorithms for classification problems on vertically partitioned data with an edge system are proposed. The effectiveness is shown by numerical simulations.