AFBS: buffer gradient selection in semi-asynchronous federated learning

Asynchronous federated learning (AFL) accelerates training by eliminating the need to wait for stragglers. However, its asynchronous nature introduces gradient staleness, where outdated gradients degrade performance. Existing solutions address this issue with gradient buffers, forming a semi-asynchronous framework. Nevertheless, this approach struggles when buffers accumulate numerous stale gradients, as blindly aggregating all gradients can harm training. To address this, we propose AFBS (Asynchronous FL Buffer Selection), the first algorithm to perform gradient selection within buffers while ensuring privacy protection. Specifically, the client sends the random projection encrypted label distribution matrix before training, and the server performs client clustering based on it. During training, server scores and selects gradients within each cluster based on their informational value, discarding low-value gradients to enhance semi-asynchronous federated learning. Extensive experiments in highly heterogeneous system and data environments demonstrate AFBS's superior performance compared to state-of-the-art methods. Notably, on the most challenging task, CIFAR-100, AFBS improves accuracy by up to 4.82 % over the previous best algorithm and reduces the time to reach target accuracy by 75.06 %. The AFBS framework. Clients first need to encrypt the label distribution using random projection before training. The server will cluster the clients based on this information. During training, the server collects gradients sent by clients until the buffer is full. Once the buffer is full, the server employs the AFBS algorithm to perform Gradient Selection on the gradients within each cluster. The remaining gradients in the buffer are then aggregated and combined once more with the global model. The server broadcasts this new global model to the clients, commencing a new round of federated learning.