Adaptive Multiple Access and Service Placement for Generative Diffusion Models

Hamidreza Mazandarani 1 , Mohammad Farhoudi 2 , Masoud Shokrnezhad 3 , and Tarik Taleb 1 · 2025

Read Adaptive Multiple Access and Service Placement for Generative Diffusion Models by Hamidreza Mazandarani 1 , Mohammad Farhoudi 2 , Masoud Shokrnezhad 3 , and Tarik Taleb 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-07978.txt.

Description

Hamidreza Mazandarani and co-authors present, in a 2025 paper, a joint plan for placing image-generating AI services on edge servers and sharing access to them. Diffusion models make rich images but need many costly steps, which strains real-time mobile use. Their framework splits the cleaning steps of generation across mixed edge machines.

A learning algorithm moves service pieces as demand shifts, while cutting steps when quality allows. A greedy access rule joins a deep learning placement method that weighs reward now and later. Sending the hidden image state between servers adds cost, which the model counts. Simulations beat fixed, single-server plans on scale and delay, the authors report.

Catalog record

CREATORHamidreza Mazandarani 1 , Mohammad Farhoudi 2 , Masoud Shokrnezhad 3 , and Tarik Taleb 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE16 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-07978

Browse the full library · About the Tower of Babel