Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PAFedFV: Personalized and Asynchronous Federated Learning for Finger Vein Recognition

Hengyu Mu, Jian Guo, Chong Han, Lijuan Sun

TL;DR

A Personalized and Asynchronous Federated Learning for Finger Vein Recognition (PAFedFV) framework, which designs personalized model aggregation method to solve the heterogeneity among non-IID data and employs an asynchronized training module for clients to utilize their waiting time.

Abstract

With the increasing emphasis on user privacy protection, biometric recognition based on federated learning have become the latest research hotspot. However, traditional federated learning methods cannot be directly applied to finger vein recognition, due to heterogeneity of data and open-set verification. Therefore, only a few application cases have been proposed. And these methods still have two drawbacks. (1) Uniform model results in poor performance in some clients, as the finger vein data is highly heterogeneous and non-Independently Identically Distributed (non-IID). (2) On individual client, a large amount of time is underutilized, such as the time to wait for returning model from server. To address those problems, this paper proposes a Personalized and Asynchronous Federated Learning for Finger Vein Recognition (PAFedFV) framework. PAFedFV designs personalized model aggregation method to solve the heterogeneity among non-IID data. Meanwhile, it employs an asynchronized training module for clients to utilize their waiting time. Finally, extensive experiments on six finger vein datasets are conducted. Base on these experiment results, the impact of non-IID finger vein data on performance of federated learning are analyzed, and the superiority of PAFedFV in accuracy and robustness are demonstrated.

PAFedFV: Personalized and Asynchronous Federated Learning for Finger Vein Recognition

TL;DR

A Personalized and Asynchronous Federated Learning for Finger Vein Recognition (PAFedFV) framework, which designs personalized model aggregation method to solve the heterogeneity among non-IID data and employs an asynchronized training module for clients to utilize their waiting time.

Abstract

With the increasing emphasis on user privacy protection, biometric recognition based on federated learning have become the latest research hotspot. However, traditional federated learning methods cannot be directly applied to finger vein recognition, due to heterogeneity of data and open-set verification. Therefore, only a few application cases have been proposed. And these methods still have two drawbacks. (1) Uniform model results in poor performance in some clients, as the finger vein data is highly heterogeneous and non-Independently Identically Distributed (non-IID). (2) On individual client, a large amount of time is underutilized, such as the time to wait for returning model from server. To address those problems, this paper proposes a Personalized and Asynchronous Federated Learning for Finger Vein Recognition (PAFedFV) framework. PAFedFV designs personalized model aggregation method to solve the heterogeneity among non-IID data. Meanwhile, it employs an asynchronized training module for clients to utilize their waiting time. Finally, extensive experiments on six finger vein datasets are conducted. Base on these experiment results, the impact of non-IID finger vein data on performance of federated learning are analyzed, and the superiority of PAFedFV in accuracy and robustness are demonstrated.
Paper Structure (22 sections, 2 theorems, 10 equations, 6 figures, 3 tables, 1 algorithm)

This paper contains 22 sections, 2 theorems, 10 equations, 6 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Finger Vein Recognition.
  4. Federated Biometrics.
  5. Proposed Method
  6. PAFedFV
  7. Personalized Model Aggregation
  8. Asynchronous Training
  9. Learning Pipeline
  10. Convergence Analysis
  11. Experiments
  12. Experimental Setup
  13. Dataset
  14. Evaluation Metrics
  15. Experimental Settings
  16. ...and 7 more sections

Key Result

Theorem 1

Meanwhile, the aggregation method proposed in this paper satisfies Then, if $p_{k}$ denotes the weight of k-th client's model, our method satisfies $\sum_{k}p_{k} = 1$.

Figures (6)

  • Figure 1: The settings of personalized FL for finger vein recognition. Our goal is to implement a personalized and efficient FL framework to adapt to the highly heterogeneous finger vein data.
  • Figure 2: The basic framework of PAFedFV. The client trains entire model through local training and uploads the federated channel $\Phi_{g}$ to server upon completing local training. On server, personalized model aggregation is performed, yielding different global models for each client. Meanwhile, clients engage in asynchronous training for local channel $\Phi_{l}$ while waiting for the server to return global model.
  • Figure 3: The comparison between PAFedFV and traditional federated learning. In traditional federated learning. The clients invest a substantial amount of time while awaiting the return of the global model from the server. In comparison, the PAFedFV employs the asynchronous training module for the client to make full use of the waiting time.
  • Figure 4: Boxplots of different dataset combinations. where (a) is the EER and TAR(@FAR=0.01) results for different datasets combinations on FV-USM (b) is on HKPU (c) is on NUPT-FV (d) is on SDUMLA (e) is on UTFVP and (f) is on VERA.
  • Figure 5: The results of comparation experiment. where (a) is the EER results (b) is the TAR@FAR=0.01.
  • ...and 1 more figures

Theorems & Definitions (2)

  • Theorem 1
  • Theorem 2