Data Collaboration Analysis Framework Using Centralization of Individual Intermediate Representations for Distributed Data Sets

Abstract

This paper proposes a data collaboration analysis framework for distributed data sets. The proposed framework involves centralized machine learning while the original data sets and models remain distributed over a number of institutions. Recently, data has become larger and more distributed with decreasing costs of data collection. Centralizing distributed data sets and analyzing them as one data set can allow for novel insights and attainment of higher prediction performance than that of analyzing distributed data sets individually. However, it is generally difficult to centralize the original data sets because of a large data size or privacy concerns. This paper proposes a data collaboration analysis framework that does not involve sharing the original data sets to circumvent these difficulties. The proposed framework only centralizes intermediate representations constructed individually rather than the original data set. The proposed framework does not use privacy-preserving computations or model centralization. In addition, this paper proposes a practical algorithm within the framework. Numerical experiments reveal that the proposed method achieves higher recognition performance for artificial and real-world problems than individual analysis.

Privacy-Preserving Data Sharing by Integrating Perturbed Distance Matrices

“…In this section, we describe one type of data collaboration analysis proposed in [13]. A schematic illustration is shown in Fig.…”

“…In [13], the following method to estimate Z and g by solving the minimal perturbation problem was proposed:…”

“…In the framework of data collaboration analysis [13], our method can be regarded as using f for mapping from the original data to the distance matrix between the original data and the anchor data, and g for mapping from the integrated graph that is expected to have the top-k nearest neighbors of all nodes to the collaboration representation.…”

“…We applied our method to two popular classification datasets in the UCI machine learning repository [9] and compared the results with those by the method proposed in [13] (called the SVD-based method). We conducted experiments where we used the Iris dataset, which contains 150 samples in a four-dimensional dataset to classify into three classes, and the Breast Cancer Wisconsin (original) dataset, which contains 699 samples in a nine-dimensional dataset to classify into two classes.…”

“…A data collaboration analysis method that safely integrates distributed data without using encryption was recently proposed [13]. Even if this method does not use cryptography, it can reduce the risk of estimating the original data.…”