PILoRA: Prototype Guided Incremental LoRA for Federated Class-Incremental Learning
About
Existing federated learning methods have effectively dealt with decentralized learning in scenarios involving data privacy and non-IID data. However, in real-world situations, each client dynamically learns new classes, requiring the global model to classify all seen classes. To effectively mitigate catastrophic forgetting and data heterogeneity under low communication costs, we propose a simple and effective method named PILoRA. On the one hand, we adopt prototype learning to learn better feature representations and leverage the heuristic information between prototypes and class features to design a prototype re-weight module to solve the classifier bias caused by data heterogeneity without retraining the classifier. On the other hand, we view incremental learning as the process of learning distinct task vectors and encoding them within different LoRA parameters. Accordingly, we propose Incremental LoRA to mitigate catastrophic forgetting. Experimental results on standard datasets indicate that our method outperforms the state-of-the-art approaches significantly. More importantly, our method exhibits strong robustness and superiority in different settings and degrees of data heterogeneity. The code is available at \url{https://github.com/Ghy0501/PILoRA}.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Federated Class-Incremental Learning | Tiny-ImageNet 10 tasks (20 classes per task) (test) | FAA74.5 | 54 | |
| Federated Class-Incremental Learning | CIFAR-100 Quantity-based label imbalance | FAA69.5 | 42 | |
| Federated Class-Incremental Learning | CIFAR-100 Distribution-based label imbalance | FAA68.5 | 39 | |
| Federated Class-Incremental Learning | CIFAR-100 | FAA (beta=0.5)76.48 | 13 | |
| Federated Class-Incremental Learning | ImageNet-R | FAA (β=0.5)53.67 | 13 | |
| Federated Continual Learning | DomainNet | Average Performance31.22 | 13 | |
| Federated Continual Learning | CIFAR-100 | Average Accuracy76.21 | 13 | |
| Federated Continual Learning | ImageNet-R | Avg Accuracy45.43 | 13 | |
| Federated Class-Incremental Learning | CIFAR-100 alpha = 6 | FAA69.5 | 5 | |
| Image Classification | ImageNet Quantity-based label imbalance 10 tasks (100 classes per task) alpha=60 | FAA68 | 4 |