Hundreds of members of the Naval Research & Development Enterprise (NR&DE), including many NAVSEA Warfare Center employees, took part in back-to-back machine learning events in February.
NUWC Newport Division hosted its first Machine Learning (ML) Summit Feb. 7-8, and Space and Naval Warfare Systems Center Pacific (SSC Pacific) hosted its second annual Workshop on Naval Applications of Machine Learning (NAML) Feb. 13-15, in San Diego.
“Working collaboratively with other Warfare Center Divisions and SYSCOMS to develop enterprise-wide expertise in machine learning will be essential to expand our advantage in this emergent and critical technology area,” said NUWC Commander, Rear Adm. Moises DelToro III.
Machine Learning Summit
Participants from NUWC Newport Division, NSWC Carderock Division, NSWC Corona Division, NSWC Crane Division, NSWC Dahlgren Division, NUWC Keyport Division and SSC Pacific gathered for the Machine Learning Summit in Rhode Island.
“The computer science discipline of machine learning has experienced rapid advancements over the past few years due to the advancements in ML algorithms, the abundance of cheap and fast computing devices, and the access to tremendous amounts of data,” according to the Machine Learning Innovation Event Guidebook. “The Navy seeks to take advantage of the ML constructs and their application to warfighting domains.”
The ML Summit had three goals: share, explore and collaborate. “This high velocity learning event was designed to bring together ML researchers from across the NAVSEA Warfare Centers and SPAWAR to discuss the state of ML at their commands, gain industry perspective, and work together to solve problems and propose collaborative projects,” said Mark Dalton, a computer scientist currently on detail to NUWC Newport Division’s Deputy Technical Director.
During the summit’s plenary session, representatives from NVIDIA addressed more than 250 employees to summarize their industry thrust areas. NVIDIA, an industry leader in the development of Artificial Intelligence (AI) and ML technology, including self-driving cars, “Smart Cities,” health care detection and diagnostics, and high-performance computing, also shared eight deep learning use-cases in federal government.
The summit exposed participants to challenges and problems as well as potential ideas and areas for collaboration. After ML concepts were pitched, participants formed teams to refine those concepts, which were later briefed to the plenary team to generate constructive comments.
Nick Manning, a computer engineer at NUWC Newport Division, attended the summit to learn more about how NUWC and other Warfare Centers are utilizing ML.
“The concept of implementing ML into the BYG-1 Cybersecurity Toolkit (CSTK), formerly the Information Assurance Toolkit, was brought to my attention in a Cybersecurity Working Group,” Manning said. “I will look into the possibility of the CSTK utilizing ML to make audit monitoring easier for the Sailors. I will also look into other potential ways the CSTK can utilize machine learning.”
As a result of the summit, the group developed a Cooperative Research and Development Agreement with NVIDIA, identified the demand to create an NR&DE-level ML/AI community of interest, and identified a top-priority collaboration: establishing an NR&DE ML Data Repository.
“We've all made some great connections that we will work to strengthen going forward,” Dalton said.
Naval Applications of Machine Learning
Dalton also was among more than 350 AI and ML subject matter experts from the DoD, academia and industry who attended the NAML workshop held the next week in San Diego. The set of methods and technologies that comprise machine learning also underlie AI, and is a key enabler to the technical vision of SSC Pacific, the Navy and the DoD.
More than 60 NAVSEA employees, most of them from the Warfare Centers, were among those who attended and presented during the three-day workshop.
Topics and capabilities discussed during the unclassified workshop included computer vision, autonomy, big data analysis, deep learning, natural language processing, and operational applications such as electronic warfare. Discussion sessions allowed attendees to meet in smaller groups to discuss their work and form collaborations in areas such as cybersecurity and biomedical research. A roundtable discussion also was held about the previous week’s ML Summit.
“We continued development of proposals and are actively pursuing means of sharing data and results with other labs,” said Michael Walton, a research scientist at SSC Pacific who attended both events.
Dr. Jennifer Williams from NUWC Keyport Division was one of several Warfare Center employees who gave talks during the NAML workshop. As a researcher for the Obsolescence Management Information System (OMIS), her presentation was focused on “Applying Machine Learning and Data Mining to Obsolescence Management.” She also attended the ML Summit at NUWC Newport Division.
“These events provide Warfare Centers with opportunities to branch out of the bubble they live in,” she said. “While they may be the main group working on a research area, playing the odds, they are not the only group exploring that area; that leads to collaboration opportunities. Similarly, showing what you have done and interacting with other researchers, regardless of both your and their research area, can greatly expand your horizons and give you a new perception on how to solve problems.”
Roger Lamb, a software developer from NSWC Dahlgren Division, gave a presentation on the Intelligent Machine Abnormality Detection and Reaction Engine (IMADRE). The focus of the project is to build a machine learning engine to determine if a file is malicious or not just by examining the binary file in a static form.
“I was surprised to see a good focus on user behavior to determine if they are doing normal activity or not,” he said. “I liked the emphasis on how machine learning can improve tasks that are time consuming or hard for a human to accomplish.”
Participants also said it was important to bring the information from the machine learning events back to their home commands and apply it to specific projects and develop capabilities.
“We need to share the information that we learned to our command as well as follow up on the sharing of our own resources to other centers that showed an interest,” said Sara Melvin, a research scientist at NSWC Port Hueneme Division, and previously at SSC Pacific. Melvin gave a presentation on event detection method with summarization — that is, identifying and summarizing occurrences, or happenings, in time-dependent text without any prior information about the events or keywords. Although she typically uses the algorithms on social media sites such as Twitter, Melvin used the data set from Fusion — part of the suite of online tools used by NAVSEA and SPAWAR — for her NAML presentation.
Adam Tucker, a computer scientist at NSWC Philadelphia Division, attended NAML to see what kind of work is being done in predictive analysis and autonomy with unmanned underwater vehicles (UUVs).
“I absorbed a lot about what type of machine learning techniques are best for different problem types, and the interesting ways to apply those techniques to different problems,” he said. “Also, I enjoyed learning about Explainable AI (XAI) and am considering its possible applications at my department.”
Dr. Katie Rainey, an SSC Pacific mathematician and co-lead for the workshop with founder Dr. Josh Harguess, said the number of attendees at this year’s workshop reflected the importance of AI/ML to industry, academia and the military, and also the keen interest of subject matter experts to network and share their findings.
"We are so proud of the overwhelming response we have gotten this year,” Rainey said. “We were able to put together an excellent technical agenda to show off a wide range of machine learning research in support of Navy needs. This allowed people to meet people with similar interests to their own or to learn about research that could support their programs."
For more news from NAVSEA, go to www.navsea.navy.mil.