The field of federated learning is moving towards addressing the challenges of heterogeneous data distributions across clients. Recent developments focus on improving the performance and fairness of federated learning models in the presence of non-IID data. Notably, researchers are exploring novel approaches to mitigate the impact of data heterogeneity, such as using hypernetworks, sheaf collaboration, and dimension-decomposed learning. These innovative methods aim to enhance the personalization and generalization of federated learning models, making them more suitable for real-world applications. Some noteworthy papers in this area include FedUHD, which proposes a federated learning framework based on hyperdimensional computing, and FedSheafHN, which introduces a sheaf collaboration mechanism for personalized subgraph federated learning. These advancements have the potential to significantly improve the effectiveness of federated learning in various domains, including computer vision, natural language processing, and graph learning.
Advances in Federated Learning for Heterogeneous Data
Sources
Mitigating Modality Quantity and Quality Imbalance in Multimodal Online Federated Learning
DFed-SST: Building Semantic- and Structure-aware Topologies for Decentralized Federated Graph Learning
Fed-Meta-Align: A Similarity-Aware Aggregation and Personalization Pipeline for Federated TinyML on Heterogeneous Data
FedUHD: Unsupervised Federated Learning using Hyperdimensional Computing
Fairness Regularization in Federated Learning
A Large-Scale Web Search Dataset for Federated Online Learning to Rank
Breaking the Aggregation Bottleneck in Federated Recommendation: A Personalized Model Merging Approach
Widening the Network Mitigates the Impact of Data Heterogeneity on FedAvg
Deploying Models to Non-participating Clients in Federated Learning without Fine-tuning: A Hypernetwork-based Approach
FedUNet: A Lightweight Additive U-Net Module for Federated Learning with Heterogeneous Models
Dextr: Zero-Shot Neural Architecture Search with Singular Value Decomposition and Extrinsic Curvature
Seeing the Many: Exploring Parameter Distributions Conditioned on Features in Surrogates
Calibrating Biased Distribution in VFM-derived Latent Space via Cross-Domain Geometric Consistency
Towards a Larger Model via One-Shot Federated Learning on Heterogeneous Client Models
Personalized Subgraph Federated Learning with Sheaf Collaboration
Trans-XFed: An Explainable Federated Learning for Supply Chain Credit Assessment
Comparison of derivative-free and gradient-based minimization for multi-objective compositional design of shape memory alloys
Learning Time-Varying Convexifications of Multiple Fairness Measures
Dimension-Decomposed Learning for Quadrotor Geometric Attitude Control with Almost Global Exponential Convergence on SO(3)
FedEve: On Bridging the Client Drift and Period Drift for Cross-device Federated Learning
Federated Distillation on Edge Devices: Efficient Client-Side Filtering for Non-IID Data
On Defining Neural Averaging
Holo-Artisan: A Personalized Multi-User Holographic Experience for Virtual Museums on the Edge Intelligence
Nonlinear Federated System Identification