FedSLS: Exploring Federated Aggregation in Saliency Latent Space
Abstract: Federated Learning (FL) is an emerging direction in distributed machine learning that enables jointly training a global model without sharing the data with server. However, data heterogeneity biases the parameter aggregation at the server, leading to slower convergence and poorer accuracy of the global model. To cope with this, most of the existing works involve enforcing regularization in local optimization or improving the model aggregation scheme at the server. Though effective, they lack a deep understanding of cross-client features. In this paper, we propose a saliency latent space feature aggregation method (FedSLS) across federated clients. By Guided BackPropagation (GBP), we transform deep models into powerful and flexible visual fidelity encoders, applicable to general state inputs across different image domains, and achieve powerful aggregation in the form of saliency latent features. Notably, since GBP is label-insensitive, it is sufficient to capture saliency features only once on each client. Experimental results demonstrate that FedSLS leads to significant improvements over the state-of-the-arts in terms of accuracies, especially in highly heterogeneous settings. For example, on CIFAR-10 dataset, FedSLS achieves 63.43% accuracy within the strongly heterogeneous environment α=0.05, which is 6% to 23% higher than the other baselines.
Primary Subject Area: [Systems] Systems and Middleware
Secondary Subject Area: [Content] Media Interpretation
Relevance To Conference: In the multimedia domain, our work on Federated Learning (FL) tackles the crucial issue of data heterogeneity that is prevalent across various applications from computer vision to medical data analysis. By developing FedSLS, we focus on optimizing aggregation weights using saliency latent variables, which enables us to capture the essence of client data more effectively. This approach not only improves model performance by harnessing the inherent learnable value of data features across clients but also emphasizes the importance of data uniqueness and relevance, leading to more robust and generalized models in multimedia applications. Our innovative use of Guided Backpropagation for data embedding stands out, as it finely sifts through data to retain only significant features, thereby enhancing the data representation for multimedia analysis.
Submission Number: 3206
Loading