Collaborative Data Analysis: Non-model Sharing-Type Machine Learning for Distributed Data

摘要

This paper proposes a novel non-model sharing-type collaborative learning method for distributed data analysis, in which data are partitioned in both samples and features. Analyzing these types of distributed data are essential tasks in many applications, e.g., medical data analysis and manufacturing data analysis due to privacy and confidentiality concerns. By centralizing the intermediate representations which are individually constructed in each party, the proposed method achieves collaborative analysis without revealing the individual data, while the learning models remain distributed over local parties. Numerical experiments indicate that the proposed method achieves higher recognition performance for artificial and real-world problems than individual analysis.