AFRL Researchers and University of Wisconsin Teams Tackle Machine Learning Challenges

AFRL Researchers and University of Wisconsin Teams Tackle Machine Learning Challenges

Agency: 
Dept. of Defense

Machine learning experts, faculty and students at the AFRL University Center of Excellence on Efficient and Robust Machine Learning, led by the University of Wisconsin-Madison, June 5, 2019. (AFRL courtesy photo)

Scientists and engineers from the Air Force Research Laboratory (AFRL) and other Department of Defense (DoD) research laboratories came together June 5 with their academic counterparts from the University of Wisconsin-Madison, the University of Chicago, and the Toyota Technological Institute at Chicago to discuss fundamental machine learning research challenges and foster new collaborations in support of the AFRL University Center of Excellence (UCoE) on Efficient and Robust Machine Learning led by the University of Wisconsin-Madison. 

The day-long event brought together over 100 machine learning government experts, faculty and students to present fundamental work in the area of machine learning (ML) and discuss unique Air Force ML challenges that support Air Force directions in artificial intelligence, autonomy, and multi-domain command and control. The event was held to mark the successful completion of the first year of a five-year effort with the University of Wisconsin-Madison MADLAB (Machine, Algorithms, and Data) team. The event included a special invited talk on machine learning and autonomy by Jean-Charles Lede, technical advisor to the AFRL commander on autonomy, a panel session on future Air Force ML challenges, and over 60 posters from both government scientists and engineers and MADLAB team members.

The objective of the UCoE, which kicked off in 2018 for five years, is to advance the fundamentals and applications of ML in Air Force applications and environments. In order to achieve this objective, new ML theory and techniques must be developed with a specific focus on unique Air Force operating challenges centered around machine learning efficiency and robustness.

To address these challenges, the UCoE is focused on four research thrusts: 1) Data Efficiency: Developing new ML theory and techniques to enable the rapid training and/or adaptation of ML models from only a small number of examples;  2) Computational Efficiency: Developing new ML systems that are operationally realizable in both hardware and software, and ML solutions that are easily scalable, from cloud-based systems to low-power devices; 3) Operational Robustness:  Developing ML algorithms that are interpretable and perform over dynamic operational conditions, missing data, and sensor failures; and 4) Adversarial Robustness: Developing new ML techniques that are resilient to adversarial attacks and data contamination.

In addition to developing the next generation of state-of-the-art machine learning techniques, the UCoE is also focused on developing the next generation of machine learning researchers and practitioners from academia and government. The UCoE has a number of ongoing and planned initiatives for collaboration, including government and academic exchanges, seminars, open challenge problems, joint research projects, and software and data collaborations 

The ERML UCoE is supported jointly by the AFRL Office of Scientific Research and the Information Directorate.  This partnership provides a pipeline for fostering and developing new fundamental techniques in machine learning and applying them to relevant problem settings and environments that the Air Force must operate in.  To learn more about the UCoE and future MADLAB events, please visit the MADLAB team’s COE website at: https://madlab.ml.wisc.edu/.  For any additional interests in coordination, please contact Dr. Lee Seversky of AFRL’s Information Directorate at lee.seversky@us.af.mil

To view the original AFRL article, visit https://www.wpafb.af.mil/News/Article-Display/Article/1913022/afrl-scientists-engineers-join-university-of-wisconsin-teams-to-address-machine/.

Category: 
Member Labs
Region: 
Northeast