SAGIPS: a physics-inspired scalable asynchronous generative inverse-problem solver

Journal Article · Wed Apr 16 00:00:00 EDT 2025 · Machine Learning: Science and Technology

DOI:https://doi.org/10.1088/2632-2153/adc8fb· OSTI ID:2557554

Lersch, Daniel ^[1]; Schram, Malachi ^[1]; Dai, Zhenyu ^[2]; Rajput, Kishansingh ^[1]; Sato, Nobuo ^[1]; Wu, Xingfu ^[3]; Childers, J. Taylor ^[3]; Goldenberg, Steven ^[1]

---

1. Thomas Jefferson National Accelerator Facility (TJNAF), Newport News, VA (United States)
2. Vixiv Inc., Cincinnati, OH (United States)
3. Argonne National Laboratory (ANL), Argonne, IL (United States)

Solving large-scale inverse problems using deep-learning algorithms have become an essential part of modern research and industrial applications. The complexity of the underlying inverse problem may require the utilization of high performance computing systems which poses a challenge on the algorithmic design of the inverse problem solver. Most deep learning algorithms require, due to their design, custom parallelization techniques in order to be resource efficient while showing a reasonable convergence. In this paper we introduce a Scalable Asynchronous Generative Inverse Problem Solver (SAGIPS) on high-performance computing systems. We present a workflow that utilizes an asynchronous ring-allreduce algorithm to transfer the gradients of the generator network across multiple GPUs. Experiments with a scientific proxy application demonstrate that SAGIPS shows near linear weak scaling, together with a convergence quality that is comparable to traditional methods. The approach presented here allows leveraging Generative Adverserial Network across multiple GPUs, promising advancements in solving complex inverse problems at scale.

• Research Organization: Thomas Jefferson National Accelerator Facility (TJNAF), Newport News, VA (United States)
• Sponsoring Organization: USDOE; USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR); USDOE Office of Science (SC), Nuclear Physics (NP)
• Grant/Contract Number: AC02-06CH11357; AC05-06OR23177; SC0023472
• OSTI ID: 2557554
• Alternate ID(s): OSTI ID: 2549470; OSTI ID: 2563545
• Report Number(s): DOE/OR/23177--7637; JLAB-CST--24-4174; arXiv:2407.00051
• Journal Information: Machine Learning: Science and Technology, Journal Name: Machine Learning: Science and Technology Journal Issue: 2 Vol. 6; ISSN 2632-2153
• Publisher: IOP Publishing
• Country of Publication: United States
• Language: English

Related Subjects

97 MATHEMATICS AND COMPUTING
deep learning
generative AI
high performance computing
inverse problem solving
scalable asynchronous workflows