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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Naibo, Deng, Yuchen, Feng, Wenjie, Yin, Jianwei, Ng, See-Kiong
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909211626045440
author Wang, Naibo
Deng, Yuchen
Feng, Wenjie
Yin, Jianwei
Ng, See-Kiong
author_facet Wang, Naibo
Deng, Yuchen
Feng, Wenjie
Yin, Jianwei
Ng, See-Kiong
contents Federated Class Incremental Learning (FCIL) is a critical yet largely underexplored issue that deals with the dynamic incorporation of new classes within federated learning (FL). Existing methods often employ generative adversarial networks (GANs) to produce synthetic images to address privacy concerns in FL. However, GANs exhibit inherent instability and high sensitivity, compromising the effectiveness of these methods. In this paper, we introduce a novel data-free federated class incremental learning framework with diffusion-based generative memory (DFedDGM) to mitigate catastrophic forgetting by generating stable, high-quality images through diffusion models. We design a new balanced sampler to help train the diffusion models to alleviate the common non-IID problem in FL, and introduce an entropy-based sample filtering technique from an information theory perspective to enhance the quality of generative samples. Finally, we integrate knowledge distillation with a feature-based regularization term for better knowledge transfer. Our framework does not incur additional communication costs compared to the baseline FedAvg method. Extensive experiments across multiple datasets demonstrate that our method significantly outperforms existing baselines, e.g., over a 4% improvement in average accuracy on the Tiny-ImageNet dataset.
format Preprint
id arxiv_https___arxiv_org_abs_2405_17457
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memory
Wang, Naibo
Deng, Yuchen
Feng, Wenjie
Yin, Jianwei
Ng, See-Kiong
Computer Vision and Pattern Recognition
Distributed, Parallel, and Cluster Computing
Machine Learning
Federated Class Incremental Learning (FCIL) is a critical yet largely underexplored issue that deals with the dynamic incorporation of new classes within federated learning (FL). Existing methods often employ generative adversarial networks (GANs) to produce synthetic images to address privacy concerns in FL. However, GANs exhibit inherent instability and high sensitivity, compromising the effectiveness of these methods. In this paper, we introduce a novel data-free federated class incremental learning framework with diffusion-based generative memory (DFedDGM) to mitigate catastrophic forgetting by generating stable, high-quality images through diffusion models. We design a new balanced sampler to help train the diffusion models to alleviate the common non-IID problem in FL, and introduce an entropy-based sample filtering technique from an information theory perspective to enhance the quality of generative samples. Finally, we integrate knowledge distillation with a feature-based regularization term for better knowledge transfer. Our framework does not incur additional communication costs compared to the baseline FedAvg method. Extensive experiments across multiple datasets demonstrate that our method significantly outperforms existing baselines, e.g., over a 4% improvement in average accuracy on the Tiny-ImageNet dataset.
title Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memory
topic Computer Vision and Pattern Recognition
Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2405.17457