Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Incentive-Aware Federated Averaging with Performance Guarantees under Strategic Participation

Fateme Maleki, Krishnan Raghavan, Farzad Yousefian

Abstract

Federated learning (FL) is a communication-efficient collaborative learning framework that enables model training across multiple agents with private local datasets. While the benefits of FL in improving global model performance are well established, individual agents may behave strategically, balancing the learning payoff against the cost of contributing their local data. Motivated by the need for FL frameworks that successfully retain participating agents, we propose an incentive-aware federated averaging method in which, at each communication round, clients transmit both their local model parameters and their updated training dataset sizes to the server. The dataset sizes are dynamically adjusted via a Nash equilibrium (NE)-seeking update rule that captures strategic data participation. We analyze the proposed method under convex and nonconvex global objective settings and establish performance guarantees for the resulting incentive-aware FL algorithm. Numerical experiments on the MNIST and CIFAR-10 datasets demonstrate that agents achieve competitive global model performance while converging to stable data participation strategies.

Incentive-Aware Federated Averaging with Performance Guarantees under Strategic Participation

Abstract

Federated learning (FL) is a communication-efficient collaborative learning framework that enables model training across multiple agents with private local datasets. While the benefits of FL in improving global model performance are well established, individual agents may behave strategically, balancing the learning payoff against the cost of contributing their local data. Motivated by the need for FL frameworks that successfully retain participating agents, we propose an incentive-aware federated averaging method in which, at each communication round, clients transmit both their local model parameters and their updated training dataset sizes to the server. The dataset sizes are dynamically adjusted via a Nash equilibrium (NE)-seeking update rule that captures strategic data participation. We analyze the proposed method under convex and nonconvex global objective settings and establish performance guarantees for the resulting incentive-aware FL algorithm. Numerical experiments on the MNIST and CIFAR-10 datasets demonstrate that agents achieve competitive global model performance while converging to stable data participation strategies.
Paper Structure (14 sections, 15 theorems, 96 equations, 2 figures, 1 algorithm)

This paper contains 14 sections, 15 theorems, 96 equations, 2 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem formulation
  3. Algorithm outline
  4. Convergence analysis
  5. Main results
  6. Analysis
  7. Numerical Results
  8. Experiment setup
  9. Observations and insights
  10. Impact of local computation ($H$)
  11. Nash equilibrium stability
  12. Conclusions
  13. Supplementary material
  14. Convex Setting

Key Result

Lemma 4.2

Consider Algorithm alg:IncentFedAvg and Definition def:main_terms. For all $k \geq 0$, we have $\bar{x}_{k+1} = \bar{x}_k - \gamma \bar{g}_k.$

Figures (2)

  • Figure 1: MNIST. Left: Global cross-entropy loss across different local steps $H$. Right: Convergence of client contributions $N_{i,r}$ toward the Nash equilibrium under IncentFedAvg.
  • Figure 2: CIFAR-10. Left: Global cross-entropy loss across different local steps $H$. Right: Convergence of client contributions $N_{i,r}$ toward the Nash equilibrium under IncentFedAvg.

Theorems & Definitions (34)

  • Definition 4.1
  • Lemma 4.2
  • Lemma 4.3
  • proof
  • Theorem 4.5: Nonconvex setting
  • Remark 4.6
  • Theorem 4.8: Convex setting
  • Lemma 4.9
  • Lemma 4.10
  • proof
  • ...and 24 more