Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FedDCL: a federated data collaboration learning as a hybrid-type privacy-preserving framework based on federated learning and data collaboration

Akira Imakura, Tetsuya Sakurai

TL;DR

A federated data collaboration learning (FedDCL), which solves communication issues by combining federated learning with recently proposed non-model share-type federated learning named as data collaboration analysis.

Abstract

Recently, federated learning has attracted much attention as a privacy-preserving integrated analysis that enables integrated analysis of data held by multiple institutions without sharing raw data. On the other hand, federated learning requires iterative communication across institutions and has a big challenge for implementation in situations where continuous communication with the outside world is extremely difficult. In this study, we propose a federated data collaboration learning (FedDCL), which solves such communication issues by combining federated learning with recently proposed non-model share-type federated learning named as data collaboration analysis. In the proposed FedDCL framework, each user institution independently constructs dimensionality-reduced intermediate representations and shares them with neighboring institutions on intra-group DC servers. On each intra-group DC server, intermediate representations are transformed to incorporable forms called collaboration representations. Federated learning is then conducted between intra-group DC servers. The proposed FedDCL framework does not require iterative communication by user institutions and can be implemented in situations where continuous communication with the outside world is extremely difficult. The experimental results show that the performance of the proposed FedDCL is comparable to that of existing federated learning.

FedDCL: a federated data collaboration learning as a hybrid-type privacy-preserving framework based on federated learning and data collaboration

TL;DR

A federated data collaboration learning (FedDCL), which solves communication issues by combining federated learning with recently proposed non-model share-type federated learning named as data collaboration analysis.

Abstract

Recently, federated learning has attracted much attention as a privacy-preserving integrated analysis that enables integrated analysis of data held by multiple institutions without sharing raw data. On the other hand, federated learning requires iterative communication across institutions and has a big challenge for implementation in situations where continuous communication with the outside world is extremely difficult. In this study, we propose a federated data collaboration learning (FedDCL), which solves such communication issues by combining federated learning with recently proposed non-model share-type federated learning named as data collaboration analysis. In the proposed FedDCL framework, each user institution independently constructs dimensionality-reduced intermediate representations and shares them with neighboring institutions on intra-group DC servers. On each intra-group DC server, intermediate representations are transformed to incorporable forms called collaboration representations. Federated learning is then conducted between intra-group DC servers. The proposed FedDCL framework does not require iterative communication by user institutions and can be implemented in situations where continuous communication with the outside world is extremely difficult. The experimental results show that the performance of the proposed FedDCL is comparable to that of existing federated learning.
Paper Structure (17 sections, 1 theorem, 23 equations, 6 figures, 3 tables)

This paper contains 17 sections, 1 theorem, 23 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Purpose and contributions
  4. Related works
  5. Federated learning
  6. Data collaboration analysis
  7. Proposal for FedDCL
  8. Basic concept
  9. Derivation
  10. Discussion on correctness
  11. Discussion on privacy
  12. Numerical evaluations
  13. Experimental conditions
  14. Experiment I: Proof-of-concept
  15. Experiment II: prediction performance for six datasets
  16. ...and 2 more sections

Key Result

Theorem 1

If the mapping functions $f^{(i)}_j$ are linear, that is, $f^{(i)}_j(X^{(i)}_j) = X^{(i)}_jF^{(i)}_j$ with $F^{(i)}_j \in \mathbb{R}^{m \times \widetilde{m}}$ and the matrices $F^{(i)}_j$ have the same range Then, for the collaboration representations $\widehat{X}$ of the FedDCL, there exist the dimensionality reduction $F$ such that

Figures (6)

  • Figure 1: A motivating example: privacy-preserving medical data analysis in situation where continuous communication with the outside world is extremely difficult for user institutions.
  • Figure 2: Concept of the proposed federated data collaboration learning (FedDCL).
  • Figure 3: Outline of the proposed FedDCL method.
  • Figure 4: Convergence history for BatterySmall. Remark: FedDCL has a higher convergence than FedAvg and DC.
  • Figure 5: Prediction performance. Note that lower RMSE and higher Accuracy mean better recognition performance. Remark: FedDCL demonstrates very high recognition performance compared to Local and comparable to FedAvg and DC.
  • ...and 1 more figures

Theorems & Definitions (2)

  • Theorem 1
  • proof