Distributed Edge Learning under Imperfect Data Sensing

2026-07-20

AI Summary

This work addresses the fundamental limitation imposed by imperfect data sensing on edge devices in distributed learning, where sensing quality is jointly influenced by modality, resolution, power, and sample size—factors often overlooked in conventional approaches that ignore the intrinsic impact of sensing noise on learning performance. The paper models sensing noise as a modality-dependent structured covariance and introduces a novel “sensing-noise–gradient alignment” criterion to replace traditional total-noise minimization. This enables joint optimization over sensing modality, resolution, power allocation, and sample count. Theoretical analysis yields convergence bounds for non-convex learning, hardware-achievable ε-stationarity limits, and a data-saturation threshold, revealing the intrinsic coupling between sensing quality and learning efficacy. Experiments demonstrate that the proposed joint optimization strategy substantially outperforms baseline methods and approaches the theoretical performance limit under stringent resource constraints.

Abstract

Distributed learning systems typically assume that local data is already available at clients with fixed quality, while in practice, data is sensed through imperfect physical processes whose quality depends on modality, resolution, sensing power, and sample size. We model sensing noise as a structured, modality-dependent covariance and derive a non-convex learning convergence bound whose irreducible sensing floor is governed by the alignment between the modality noise covariance and the loss-sensitivity geometry. Thus, the optimal modality minimizes this noise-gradient alignment rather than total noise power alone. The analysis further yields a sensor-hardware achievability bound for epsilon-stationarity and a hardware-saturation threshold on the accumulated dataset size. We jointly optimize modality, resolution, power, and sample count and demonstrate the performance gain through simulations.

Problem

• distributed edge learning
• imperfect data sensing
• modality-dependent noise
• sensing quality
• non-convex convergence

Innovation

• imperfect data sensing
• modality-dependent noise
• non-convex convergence bound
• noise-gradient alignment
• hardware-achievable learning