Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Owen Sampling Accelerates Contribution Estimation in Federated Learning

Hossein KhademSohi, Hadi Hemmati, Jiayu Zhou, Steve Drew

TL;DR

This paper tackles the challenge of fairly attributing client contributions in Federated Learning without incurring the prohibitive cost of exact Shapley value computation. It introduces FedOwen, which combines Owen sampling with an $\eta$-truncation to produce unbiased Shapley estimates under a fixed valuation budget, and couples this with an adaptive $\epsilon$-greedy client selector to maintain diversity and accelerate convergence in non-IID settings. A contribution-weighted aggregation scheme further biases the global model toward high-value clients, improving final accuracy. Empirical results across multiple datasets and heterogeneity configurations show FedOwen achieving up to 23% higher final accuracy than strong baselines, demonstrating scalable, fair, and efficient contribution estimation for FL.

Abstract

Federated Learning (FL) aggregates information from multiple clients to train a shared global model without exposing raw data. Accurately estimating each client's contribution is essential not just for fair rewards, but for selecting the most useful clients so the global model converges faster. The Shapley value is a principled choice, yet exact computation scales exponentially with the number of clients, making it infeasible for large federations. We propose FedOwen, an efficient framework that uses Owen sampling to approximate Shapley values under the same total evaluation budget as existing methods while keeping the approximation error small. In addition, FedOwen uses an adaptive client selection strategy that balances exploiting high-value clients with exploring under-sampled ones, reducing bias and uncovering rare but informative data. Under a fixed valuation cost, FedOwen achieves up to 23 percent higher final accuracy within the same number of communication rounds compared to state-of-the-art baselines on non-IID benchmarks.

Owen Sampling Accelerates Contribution Estimation in Federated Learning

TL;DR

This paper tackles the challenge of fairly attributing client contributions in Federated Learning without incurring the prohibitive cost of exact Shapley value computation. It introduces FedOwen, which combines Owen sampling with an -truncation to produce unbiased Shapley estimates under a fixed valuation budget, and couples this with an adaptive -greedy client selector to maintain diversity and accelerate convergence in non-IID settings. A contribution-weighted aggregation scheme further biases the global model toward high-value clients, improving final accuracy. Empirical results across multiple datasets and heterogeneity configurations show FedOwen achieving up to 23% higher final accuracy than strong baselines, demonstrating scalable, fair, and efficient contribution estimation for FL.

Abstract

Federated Learning (FL) aggregates information from multiple clients to train a shared global model without exposing raw data. Accurately estimating each client's contribution is essential not just for fair rewards, but for selecting the most useful clients so the global model converges faster. The Shapley value is a principled choice, yet exact computation scales exponentially with the number of clients, making it infeasible for large federations. We propose FedOwen, an efficient framework that uses Owen sampling to approximate Shapley values under the same total evaluation budget as existing methods while keeping the approximation error small. In addition, FedOwen uses an adaptive client selection strategy that balances exploiting high-value clients with exploring under-sampled ones, reducing bias and uncovering rare but informative data. Under a fixed valuation cost, FedOwen achieves up to 23 percent higher final accuracy within the same number of communication rounds compared to state-of-the-art baselines on non-IID benchmarks.

Paper Structure

This paper contains 33 sections, 11 equations, 10 figures, 3 tables, 7 algorithms.

Table of Contents

  1. Related Work
  2. Communication and Computation Efficiency
  3. Valuation Precision Algorithms
  4. Robustness and Security
  5. Data-Free Contribution Estimation
  6. Incentive- and Bandit-Driven Client Selection
  7. Multi-Armed Bandits for Client Scheduling
  8. Methodology
  9. Background
  10. Owen Sampling with Early Truncation
  11. Convergence Analysis of Owen Sampling
  12. Step 1 - Integral representation.
  13. Step 2 - Monte Carlo estimate at each $q$.
  14. Step 3 - Riemann approximation of the integral.
  15. Step 4 - Two-stage limit.
  16. ...and 18 more sections

Figures (10)

  • Figure 1: System workflow of the proposed adaptive FL framework
  • Figure 2: Model performance over rounds with imbalance 0.01 and Dirichlet $\alpha$ = 0.01, averaged across seeds.
  • Figure 3: Effect of inclusion levels (Q) on Final Accuracy
  • Figure 4: Effect of $\epsilon$ on Final Accuracy
  • Figure 5: MNIST class counts before and after applying imbalance factors 0.1, 0.05, and 0.01.
  • ...and 5 more figures