This work revisits client drift from a novel frequency-domain perspective and uncovers a critical Spectral Bias of Drift: inter-client gradient divergence is predominantly concentrated in low-frequency components which encode client-specific distributional shifts, while high-frequency components representing fine-grained features remain relatively consistent.
Abstract
Federated Learning (FL) is fundamentally challenged by statistical heterogeneity, where non-identically distributed (non-IID) data induces client drift that severely hampers global convergence. While existing approaches attempt to mitigate this drift through spatial-domain gradient correction or regularization, they overlook the intrinsic spectral structure of optimization signals. In this work, we revisit client drift from a novel frequency-domain perspective and uncover a critical Spectral Bias of Drift: inter-client gradient divergence is predominantly concentrated in low-frequency components which encode client-specific distributional shifts, while high-frequency components representing fine-grained features remain relatively consistent. Motivated by this, we propose SpecGradFilter, a unified Spectral Gradient Filtering Framework that tames heterogeneity by suppressing discordant low-frequency signals. Crucially, we demonstrate that SpecGradFilter is a generalizable principle, effective not only via precise FFT-based truncation but also through spatial approximations like Gaussian detrending. Extensive experiments on benchmarks such as CIFAR-10/100 and Tiny-ImageNet demonstrate that SpecGradFilter significantly performs better performance in highly Non-IID settings with negligible communication overhead, establishing a new paradigm for robust federated optimization.
This paper proposes Federated Learning with Consistency Optimization Algorithms (FedCO), a novel optimization framework that incorporates a label-skew-aware correction loss and neural feature distribution regularization during local training that significantly improves accuracy and convergence under diverse non-IID settings.
Ruiqi Wu, Yehong Li, Hongjie Guo et al.· Computers, Materials & C...· 0 citations
This work comprehensively considering the effects of client drift during the training process, and quantifying it as the aggregation error, proposes adaptive gradient regularization, which is based on gradient regularization and further and applies different regularization strengths to each parameter based on the magnitude of the parameter variance between the local model and the global model.
Xiang Wang, Lei Tian, Jiahao Gan et al.· Neural Networks· 0 citations
Federated learning enables privacy-preserving distributed training but suffers from client drift and slow convergence under statistical data heterogeneity. Most existing federated optimization methods address client drift via parameter-space constraints or aggregation-level corrections, while fewer works directly shape the gradient covariance spectral structure of the optimization landscape. This paper analyzes the convergence problem from a spectral perspective, revealing that non-IID data causes spectral diffusion in the gradient covariance matrix and degrades convergence. Guided by random matrix theory, we propose federated spectral regularization (Fed-SR), a computationally efficient method that indirectly constrains spectral spread via gradient norm regularization. Although computing the regularizer gradient requires Hessian vector products, our optimized auto-differentiation implementation avoids storing full Hessian matrices and restricts extra computational overhead to a negligible level. Experiments on CIFAR-10, CIFAR-100, and other benchmarks show that Fed-SR outperforms baselines including FedAvg, FedProx, and SCAFFOLD in non-IID scenarios, reducing communication rounds and improving accuracy and stability. Ablation studies, spectral analysis, and controlled spectral feature manipulation experiments provide consistent empirical evidence showing a strong empirical association between the “spectral concentration” effect and performance gains, offering mechanistic interpretability consistent with our proposed theoretical framework within the tested experimental settings.
Polaris is a client selection framework for asynchronous and heterogeneous federated learning that employs quality-aware sampling to accelerate convergence. Its reliance on scalar update norms as quality indicators exposes it to exploitation, in which clients submit near-zero Gaussian updates that inflate the update magnitude proxy within the Polaris sampling objective without contributing meaningful gradient information. This work introduces a lightweight reliability-aware extension that embeds two server-side signals, Exponential Moving Average (EMA)-based magnitude credibility and cosine similarity-based directional alignment, directly into the staleness and aggregation weight proxy computation without modifying the geometric programming objective or aggregation semantics. Experiments across ResNet-18 on CIFAR-10, LeNet-5 on FEMNIST, and LeNet-5 on MNIST under non-IID data distributions confirm that attack effectiveness and defense discrimination are governed by task complexity and non-IID (non-Independent and Identically Distributed) degree. The proposed method demonstrates partial resistance to convergence collapse at lower adversarial ratios while introducing only a marginal accuracy overhead under clean conditions.
A. Wicaksana, Bayu Anggorojati, M. E. Aminanto et al.· International Conference on...· 0 citations
TRISHUL is proposed, a spectral-control framework for robust federated PEFT that follows the FL no-raw-data-sharing setting but does not itself provide formal privacy guarantees, and improves convergence, stability, and final performance over federated LoRA baselines.
SensCluster is proposed, a novel sensitivity-aware CFL framework that constructs compact client representations by selecting parameters that are most responsive to local feature distributions, and consistently outperforms state-of-the-art CFL methods across diverse feature skew scenarios.
Jiaqi Wang, Tobias Schlagenhauf, Setareh Maghsudi· Proceedings of the 32nd ACM...· 0 citations
A weeklong summer workshop brought higher education faculty to campus to explore how AI and machine learning materials can be adapted for their classrooms.
