ECP Podcast interviews Argonne’s Logan Ward on ExaLearn progress
June 17, 2022 — In this latest episode of the Let’s Talk Exascale podcast, Scott Gibson interviews Logan Ward of Argonne National Laboratory. The podcast was released on June 16, 2022.
Hi and welcome. This is where we explore the efforts of the Department of Energy’s (DOE) Exascale Computing Project (ECP), from development challenges and achievements to the expected ultimate impact of exascale computing on society.
As the prevalence of data, computational methods, and the power of computers have taken off in recent years, so has machine learning, or ML, a subset of artificial intelligence. With ML, computers, software, and electronic devices can function in a way similar to the human brain, perform natural language processing, make cars self-driving, personalize our news feeds, and a myriad of other conveniences.
Additionally, ML technologies could be profoundly important in the field of computational and experimental science and engineering, where breakthroughs that change our lives occur. The ML will certainly play a role in the research that will be carried out on the first exascale supercomputers deployed by the DOE.
Not only do ML technologies create inspiring new opportunities for scientific discovery, but they also hold promise for the design and use of exascale computers themselves. HPC for ML and ML for HPC seem to be on the horizon.
ECP launched the ExaLearn co-design center in 2018 to leverage ML. The primary focus of ExaLearn is to provide exascale ML software for ECP applications, other ECP co-design centers, DOE experimental facilities, and advanced computing facilities.
ExaLearn, a collaboration of several DOE laboratories, is led by Frank Alexander, deputy director of the Computational Science Initiative at Brookhaven National Laboratory. An article entitled “Co-design Center for Exascale Machine Learning Technologies (ExaLearn” in The international journal of high performance computing applications by Alexander and the ExaLearn Contributors offers a detailed view of what ExaLearn is all about.
Our guest in this episode, Argonne National Laboratory materials scientist Logan Ward, is a researcher with the ExaLearn team. ExaLearn activities are varied. Logan’s team is working on smart workflows that use AI to decide which simulations to run. Examples of other types of ExaLearn research are making surrogates to operate instead of experiments and simulations, and reinforcement learning, which trains models to make a sequence of decisions.
We’ll get Logan’s perspective on the following: the general context of the intersection of ML and HPC, the different groups and main objectives of ExaLearn, the specific work Logan is involved in, challenges, successes, etc.
“The lines between what machine learning is and what conventional HPC is have blurred a lot. There are ways to use machine learning to tell you which simulations you should actually run. There are ways to integrate machine learning directly into simulation codes. And, all of this has really changed what it is to envision the high performance computing applications of the future. There will likely be machine learning at some level, and where ExaLearn comes in is to figure out how it’s going to go in a way that the whole HPC community can take into account. —Logan Ward, science assistant at Argonne National Laboratory.
(4:09) The general context of the link between ML and HPC
(5:35) A brief description of the ExaLearn co-design center
(6:50) Logan’s story with ExaLearn
(7:57) What the ExaLearn team that Logan is part of is working on
(9:52) The challenges the team faced
(12:29) Team successes
(13:45) The implications of what has been accomplished
(15:25) What about the exascale that excites Logan
(17:02) The next steps
Source: Exascale Calculation Project