Using Bayesian Techniques to Speed Up HPC Solutions

Bayesian inference interprets probability as a measure of believability or confidence that an individual may possess about the occurrence of a particular event.

September 07, 2021 /

Julia Chamon

 1s and 0s on an abstract background

Recent advances at IBM Research in the area of Bayesian statistical techniques have resulted in the development of an IBM software product that is compatible with any High-Performance Computing (HPC) vendor system and which dramatically speeds up the total time it takes to complete hundreds and thousands of iterative HPC jobs. Bayesian Statistics is a branch of statistics that provides tools which help in understanding the probability of the occurrence of an event with respect to the new data introduced. In particular, Bayesian inference interprets probability as a measure of believability or confidence that an individual may possess about the occurrence of a particular event.

On Thursday, August 26, IBM’S Terry Leatherland and Cheruki Choudary, PhD., discussed IBM’S Bayesian Optimization Accelerator (BOA) in detail. During the seminar, Leatherland introduced Bayesian Statistics, discussed realistic outcomes BOA-AI intelligence, and spoke on current areas of focus, such as Reservoir Simulation, Computational Fluid Dynamics, Computational Biology/Chemistry and Drug Discovery. According to Choudary, other potential areas of research may come forth as the science is tested. It doesn’t matter what HPC system is used, as the software AI engine is out of band. Results for tuning and optimizing the “next run” are created in seconds with API feedback to the associated simulation solver for the “next run.”

In creating the software, IBM Research has worked with scientists all over the world ranging from UT Austin Petroleum Engineering to Merck Pharmaceuticals in the UK to perfect the methodology. The software is free to the academic community for the purpose of  research, while it is a paid subscription for commercial/industrial entities.

Terry Leatherland is a resident of Sugar Land, Texas. As a graduate of the University of Alabama, BS Chemistry/Computer Science, his work for the first 15 years spanned the Chemistry Quality Assurance environment for HP across the Gulf Coast. As his career progressed, Terry transitioned to HP as an US Enterprise Systems Architect for Oracle, PeopleSoft, SAP and other enterprise systems. After 15 years at HP, Terry joined IBM as a cross-brand Systems Architect. He currently serves as a Client Technical Architect for the states of Texas, Illinois and Ohio. His role includes healthcare, education, life sciences and government entities. He also serves as a go-to on High Performance Computing, AI servers and storage solutions for IBM across the U.S.

Cheruki Choudary is an AI Specialist for IBM in Austin, TX. Choudary is a graduate of the Jawaharlal Nehru Technological University, BT, and the University of South Carolina-Colombia He is a multifaceted computer scientist, researcher and technologist. With experience in multiple areas of computing including Artificial Intelligence, High Performance Computing, numerical simulations, multimedia processing/retrieval, and hardware design.

News Category
Institute Happenings
UH Data Science News
Research Topics