SFedCA: Credit Assignment-Based Active Client Selection Strategy for Spiking Federated Learning

TL;DR

This work tackles non-IID challenges in spiking federated learning by introducing SFedCA, a credit assignment-based active client selection method that leverages firing-rate differences to balance the global data distribution. By evaluating firing-rate changes before and after local training, SFedCA identifies clients that contribute new distributional information and selectively aggregates them, improving both convergence speed and accuracy. Across MNIST, Fashion-MNIST, and CIFAR-10 under multiple non-IID schemes, SFedCA consistently outperforms random selection and existing active methods, with fewer communication rounds and noticeably lower energy costs. The results indicate SFedCA's practical potential for energy-efficient, privacy-preserving distributed learning on resource-constrained devices, especially in heterogeneous data environments.

Abstract

Spiking federated learning is an emerging distributed learning paradigm that allows resource-constrained devices to train collaboratively at low power consumption without exchanging local data. It takes advantage of both the privacy computation property in federated learning (FL) and the energy efficiency in spiking neural networks (SNN). Thus, it is highly promising to revolutionize the efficient processing of multimedia data. However, existing spiking federated learning methods employ a random selection approach for client aggregation, assuming unbiased client participation. This neglect of statistical heterogeneity affects the convergence and accuracy of the global model significantly. In our work, we propose a credit assignment-based active client selection strategy, the SFedCA, to judiciously aggregate clients that contribute to the global sample distribution balance. Specifically, the client credits are assigned by the firing intensity state before and after local model training, which reflects the local data distribution difference from the global model. Comprehensive experiments are conducted on various non-identical and independent distribution (non-IID) scenarios. The experimental results demonstrate that the SFedCA outperforms the existing state-of-the-art spiking federated learning methods, and requires fewer communication rounds.

Figures (8)

• Figure 1: The influence of selecting different clients. In case 1, client 1 and 2 participate in the aggregation; in case 2, client 2 and 3 participate.
• Figure 2: Information processing in IF neuron model. The neuron receives input spikes causing the change of membrane potential, and when it exceeds the threshold, the neuron fires a spike and resets the membrane potential.
• Figure 3: The framework of SFedCA. In the local training process, the client calculates the firing rate difference $\Delta \boldsymbol{R}_k^r$ according to the global model $\boldsymbol{W}^r$ and the updated local model $\boldsymbol{W}_k^{r+1}$. The server selects $P$ clients with higher firing rate differences from client candidates $\mathbb{S}$, and gets the new global model by aggregating the parameters of selected clients.
• Figure 4: The convergence curves of SFedCA and other spiking FL methods on MNIST with different client numbers.
• Figure 5: The performance of SFedCA with different selected client numbers.