Training Diffusion Models with Federated Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: de Goede, Matthijs, Cox, Bart, Decouchant, Jérémie
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916293201887232
author de Goede, Matthijs
Cox, Bart
Decouchant, Jérémie
author_facet de Goede, Matthijs
Cox, Bart
Decouchant, Jérémie
contents The training of diffusion-based models for image generation is predominantly controlled by a select few Big Tech companies, raising concerns about privacy, copyright, and data authority due to their lack of transparency regarding training data. To ad-dress this issue, we propose a federated diffusion model scheme that enables the independent and collaborative training of diffusion models without exposing local data. Our approach adapts the Federated Averaging (FedAvg) algorithm to train a Denoising Diffusion Model (DDPM). Through a novel utilization of the underlying UNet backbone, we achieve a significant reduction of up to 74% in the number of parameters exchanged during training,compared to the naive FedAvg approach, whilst simultaneously maintaining image quality comparable to the centralized setting, as evaluated by the FID score.
format Preprint
id arxiv_https___arxiv_org_abs_2406_12575
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Training Diffusion Models with Federated Learning
de Goede, Matthijs
Cox, Bart
Decouchant, Jérémie
Machine Learning
Distributed, Parallel, and Cluster Computing
I.2.11
The training of diffusion-based models for image generation is predominantly controlled by a select few Big Tech companies, raising concerns about privacy, copyright, and data authority due to their lack of transparency regarding training data. To ad-dress this issue, we propose a federated diffusion model scheme that enables the independent and collaborative training of diffusion models without exposing local data. Our approach adapts the Federated Averaging (FedAvg) algorithm to train a Denoising Diffusion Model (DDPM). Through a novel utilization of the underlying UNet backbone, we achieve a significant reduction of up to 74% in the number of parameters exchanged during training,compared to the naive FedAvg approach, whilst simultaneously maintaining image quality comparable to the centralized setting, as evaluated by the FID score.
title Training Diffusion Models with Federated Learning
topic Machine Learning
Distributed, Parallel, and Cluster Computing
I.2.11
url https://arxiv.org/abs/2406.12575