DOLFIN: Balancing Stability and Plasticity in Federated Continual Learning
This addresses the challenge of efficient and stable learning in privacy-preserving federated continual learning, though it appears incremental as it builds on existing techniques like LoRA and ViTs.
The paper tackles the problem of balancing performance, privacy, and communication efficiency in federated continual learning by introducing DOLFIN, a method combining Vision Transformers with low-rank adapters and DualGradient Projection Memory. The result shows DOLFIN consistently surpasses six baselines in final average accuracy on datasets like CIFAR-100 and ImageNet-R while matching their memory footprint.
Federated continual learning (FCL) enables models to learn new tasks across multiple distributed clients, protecting privacy and without forgetting previously acquired knowledge. However, current methods face challenges balancing performance, privacy preservation, and communication efficiency. We introduce a Distributed Online LoRA for Federated INcremental learning method DOLFIN, a novel approach combining Vision Transformers with low-rank adapters designed to efficiently and stably learn new tasks in federated environments. Our method leverages LoRA for minimal communication overhead and incorporates DualGradient Projection Memory (DualGPM) to prevent forgetting. Evaluated on CIFAR-100, ImageNet-R, ImageNet-A, and CUB-200 under two Dirichlet heterogeneity settings, DOLFIN consistently surpasses six strong baselines in final average accuracy while matching their memory footprint. Orthogonal low-rank adapters offer an effective and scalable solution for privacy-preserving continual learning in federated settings.