Joint Optimization of Resource Allocation and Data Selection for Fast and Cost-Efficient Federated Edge Learning

TL;DR

This work addresses FEEL under constrained wireless resources and data mislabeling by formulating a joint optimization of resource allocation and data selection. It derives a one-round convergence upper bound and transforms the problem into a tractable form, then decomposes it into a resource-allocation subproblem solved by a matching-based algorithm with CCP-based power control and a data-selection subproblem solved via gradient projection and binary recovery. The proposed suboptimal algorithms demonstrate substantial improvements in convergence speed and net cost on MNIST and Fashion-MNIST compared with baselines, highlighting practical gains in fast, cost-efficient FEEL. The approach offers a principled framework for jointly managing communication and data quality aspects at the edge, with potential extensions to multi-task and cross-domain scenarios.

Abstract

Deploying federated learning at the wireless edge introduces federated edge learning (FEEL). Given FEEL's limited communication resources and potential mislabeled data on devices, improper resource allocation or data selection can hurt convergence speed and increase training costs. Thus, to realize an efficient FEEL system, this paper emphasizes jointly optimizing resource allocation and data selection. Specifically, in this work, through rigorously modeling the training process and deriving an upper bound on FEEL's one-round convergence rate, we establish a problem of joint resource allocation and data selection, which, unfortunately, cannot be solved directly. Toward this end, we equivalently transform the original problem into a solvable form via a variable substitution and then break it into two subproblems, that is, the resource allocation problem and the data selection problem. The two subproblems are mixed-integer non-convex and integer non-convex problems, respectively, and achieving their optimal solutions is a challenging task. Based on the matching theory and applying the convex-concave procedure and gradient projection methods, we devise a low-complexity suboptimal algorithm for the two subproblems, respectively. Finally, the superiority of our proposed scheme of joint resource allocation and data selection is validated by numerical results.

Figures (6)

• Figure 1: System model. Device $k$ has $|\mathcal{D}_k| = 6$ images, two of which are mislabeled. The text above each image represents the corresponding label, with black text and red text representing correct and incorrect labels, respectively.
• Figure 2: A schematic diagram for solving Problem \ref{['EqualProblem']}.
• Figure 3: The convergence curves of Algorithm \ref{['algpowall']} under different initial points.
• Figure 4: Convergence and cumulative net cost of FEEL under different schemes.
• Figure 5: Effect of the mislabeled proportion.

Theorems & Definitions (5)

• Lemma 1
• Lemma 2
• Lemma 3
• Definition 1: Matching Model
• Lemma 4