Inaugural industry forum inspires ML community


September 22, 2021 – The Lawrence Livermore National Laboratory held its first-ever Machine Learning Industry Forum (ML4I) from August 10 to 12. Co-organized by the High Performance Computing Innovation Center (HPCIC) and the Data Science Institute (DSI) of the Lab, the virtual event brought together more than 500 participants from the Department of Energy (DOE) complex, commercial companies, corporations professionals and universities. Industry sponsors included ArcelorMittal, Cerebras Systems, Ford Motor Company, IBM, Intel, SambaNova Systems, NVIDIA, Intersect360 Research and Rhino Health.

The objectives of the forum were to encourage and elucidate the adoption of machine learning (ML) methods for practical results in various industries, especially manufacturing. Discussions, panels and presentations were organized around three high level themes: industrial applications, tools and techniques and the impact and potential of ML in industry.

“The forum was created based on the interest and experience of LLNL to help the industry develop artificial intelligence [AI] and ML tools for applications such as manufacturing, ”said Wayne Miller, Acting Director of HPCIC. In addition to the HPCIC, which encourages IT collaborations with the private sector, the lab’s industry-focused efforts include the High Performance Computing for Energy Innovation (HPC4EI) program and its manufacturing (HPC4Mfg) and materials ( HPC4Mtls), which leverage DOE’s HPC facilities to improve energy efficiency and streamline materials development and manufacturing processes. Additionally, LLNL’s Office of Innovation and Partnerships (IPO) engages with industry to drive economic growth and negotiates business licenses and cooperative agreements.

Miller explained, “There is a need to develop collaborations between our data scientists who ‘know how to make ML work’ and industry users who have data, but not a lot of experience in developing data tools. ML. CIO Director Michael Goldman added, “CIO’s research and outreach efforts complement HPCIC’s IT resources and expertise. It made sense to join forces on this forum.

Brenda Ng, LLNL data scientist who co-hosted the event, said: “My daily work is focused on research and deployment projects. I love applied research, so the forum gave me the opportunity to hear the experiences and solutions of others. It was also an outreach opportunity to help industry contacts understand what the Lab does. “

Main variety

Speakers from industry, academia and government took turns initiating the agenda. Ford Motor Company’s Devesh Upadhyay described ML and data-driven approaches for various aspects of vehicle design and manufacture, including surrogate models and physics-based neural networks. Pamela Isom, Director of the Office of Artificial Intelligence Technology (AITO) at DOE, highlighted the importance of improving AI / ML reliability and cybersecurity risk management and provided an overview from AITO.

Pieter Abbeel of the Robot Learning Lab at the University of California at Berkeley presented strategies for developing neural networks pre-trained in robotics. The Robot Learning Lab studies how to make existing AI systems smarter and how AI can advance science and engineering.

A robot’s brain is a network of neurons trained to perform tasks that it learns from images, text, simulations, demonstrations, and other data. Abbeel discussed different types of learning in this context, including multitasking reinforcement learning (RL), unsupervised representational learning, few stroke imitation learning, and human RL in the loop. “I was excited to share some of the latest advancements in AI robotics with a wider audience, as well as my vision for future research needed in space,” he said.

Capacity overview

The event covered the role of HPC, with presentations on ML, RL compute workflows for simulations and cognitive simulation. Industry use cases presented to the public included AI for inspection of defects in steel, computer vision, and image processing techniques to automate quality control processes, workflows. convergent HPC and AI work for drug discovery, ML to predict cardiac response to a mitral valve device, and uncertainty quantification and surrogate modeling in carbon dioxide capture systems.

LLNL speakers described areas in which ML and related data sciences are impacting the lab, such as predicting the strength and performance of materials and optimizing manufacturing processes. Computer engineer Vic Castillo presented the results of some of his HPC4Mfg projects, which use simulations of critical and energy-intensive manufacturing processes to generate data for ML routines. He said, “The forum was a great platform to showcase the wide variety of ML capabilities in the lab to a wider industry audience. “

The Castillo team developed fast predictive models of computationally expensive simulations that partners can run on gaming desktops. He explained, “This gives the production engineer good real-time forecasting. to avoid wasting energy and producing poor quality products. Large-scale simulations can be expensive for private industry, Castillo noted, so “we have to be careful to get the most useful information from the smallest number of simulations.”

Wisdom panel

The forum consisted of two panel sessions. The first discussed opportunities for recruiting and training ML talent, integrating them into the workforce, and providing resources to develop AI / ML tools. Part of this effort is to bridge the gap between ML taught in the classroom and its practical application in the real world – a goal of the Data Science Summer Institute and other student internship programs around the Lab.

Goldman, who moderated the first panel, said: “We felt it was crucial to have a conversation with the employees at ML4I. National laboratories and commercial enterprises can offer students and recent graduates practical opportunities to develop and apply their skills. As employers, we gain staff who are passionate about advancing AI / ML and related fields. “

The second panel session looked at the legal, ethical and cost-benefit challenges of sharing and securing datasets – for example, facial recognition or open source image collection. According to Miller: “Both public and private institutions need to tackle these issues, as data is the primary resource for any AI / ML development. Ng added, “This panel opened my eyes to data security and its relevance to the lab by balancing data sharing and protecting customer privacy. “

Common threads

Overall, Goldman said, “The event highlighted some commonalities between the participating organizations, and we could have spent over three days discussing ways to work together. About 50 presenters responded to the call for abstracts in April, and the high attendance at the event and the response from the audience make it a ML4I will Forum probably next year.

“The pandemic can make it more difficult to forge partnerships with industry,” noted Ng, who has a new appreciation for the way online events are handled. “I think the live and interactive nature of our event, rather than asking speakers to provide pre-recorded video, was more appealing to potential speakers and the audience. “

Besides Goldman, Miller and Ng, the ML4The organizing committee included Philip Cameron-Smith and AJ Simon, the HPCIC staff and group leaders in the Physical and Life Sciences Branch of the LLNL; Charity Follett, Head of IPO Business Development; and directors Rosie Aguilar, Florann Mahler and Katie Thomas.


Source: Lawrence Livermore National Laboratory


Sam D. Gomez