Abstract:Federated Class Incremental Learning (FCIL) is a new direction in continual learning (CL) for addressing catastrophic forgetting and non-IID data distribution simultaneously. Existing FCIL methods call for high communication costs and exemplars from previous classes. We propose a novel rehearsal-free method for FCIL named prototypes-injected prompt (PIP) that involves 3 main ideas: a) prototype injection on prompt learning, b) prototype augmentation, and c) weighted Gaussian aggregation on the server side. Our experiment result shows that the proposed method outperforms the current state of the arts (SOTAs) with a significant improvement (up to 33%) in CIFAR100, MiniImageNet and TinyImageNet datasets. Our extensive analysis demonstrates the robustness of PIP in different task sizes, and the advantage of requiring smaller participating local clients, and smaller global rounds. For further study, source codes of PIP, baseline, and experimental logs are shared publicly in <a class="link-external link-https" href="https://github.com/anwarmaxsum/PIP" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges faced in Federated Class Incremental Learning (FCIL) in the federated learning environment, specifically including: 1. **Catastrophic Forgetting**: While continuously learning new classes, the model is prone to forget the classes it has learned before. This is especially prominent in a dynamic data environment because the dataset will introduce new classes as time changes. 2. **Non - IID (Non - Independent and Identically Distributed) Data**: Data from different clients may have different distribution characteristics, making it difficult to reach the global optimum during model training. 3. **High Communication Cost**: Existing FCIL methods usually need to transmit a large number of parameters, which not only increases the communication cost but also prolongs the training time. 4. **Data Privacy**: In federated learning, protecting data privacy is an important issue. Existing methods protect privacy by sharing perturbed images, but this may leak information. To solve these problems, the paper proposes a new method named "Prototypes - Injected Prompt" (PIP). The main contributions and innovations of this method include: - **Prompt Learning**: The client only updates and shares a small - scale trainable parameter (called a prompt), while the backbone network remains frozen. This can significantly reduce the communication cost and training time. - **Shared Prototypes**: By sharing prototypes among clients, the problem of non - IID data is solved, and data privacy is guaranteed. Each client can enhance its local training effect by injecting shared prototypes. - **Prototype Enhancement**: To deal with the class - imbalance problem, a prototype enhancement mechanism is proposed to generate more prototypes to balance the number of samples in different classes. - **Weighted Gaussian Aggregation**: On the server side, the weighted Gaussian aggregation method is used to adjust the contribution of clients to the global model according to their participation and the number of training samples, thereby improving the generalization ability of the global model. Through these innovations, the PIP method has achieved performance significantly better than existing methods on multiple benchmark datasets, especially on the CIFAR100, MiniImageNet, and TinyImageNet datasets.

PIP: Prototypes-Injected Prompt for Federated Class Incremental Learning

Federated Class-Incremental Learning with Prompting

FPPL: An Efficient and Non-IID Robust Federated Continual Learning Framework

Multimodal Parameter-Efficient Few-Shot Class Incremental Learning

Learning Prompt with Distribution-Based Feature Replay for Few-Shot Class-Incremental Learning

Pseudo Initialization Based Few-Shot Class Incremental Learning

TARGET: Federated Class-Continual Learning Via Exemplar-Free Distillation

A Cognition-Driven Framework for Few-Shot Class-Incremental Learning

Few Shot Class Incremental Learning via Efficient Prototype Replay and Calibration

Few-Shot Class-Incremental Learning Via Class-Aware Bilateral Distillation

FineFMPL: Fine-grained Feature Mining Prompt Learning for Few-Shot Class Incremental Learning

FedProK: Trustworthy Federated Class-Incremental Learning via Prototypical Feature Knowledge Transfer

Class-Incremental Exemplar Compression for Class-Incremental Learning

Privacy-Preserving Federated Class-Incremental Learning

Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memory

Rehearsal-free Federated Domain-incremental Learning

Overcoming Catastrophic Forgetting in Federated Class-Incremental Learning via Federated Global Twin Generator

Improved Continually Evolved Classifiers for Few-Shot Class-Incremental Learning

Federated Class-Incremental Learning with Hierarchical Generative Prototypes

Federated Class-Incremental Learning with New-Class Augmented Self-Distillation

Prompting to Prompt for Rehearsal-Free Class Incremental Learning.