Research for Generalizable, Explainable and Robust Machine Learning
Abstract
We are requesting instrumentation at Rutgers UniversityÕs ``Center for Computational Biomedicine, Imaging and Modeling (CBIM) which is directed by Dr. Metaxas and focuses on research in novel machine learning, physics-based modeling, computer vision, AI, human behavior analytics, visualization and medical image analytics. The requested instrumentation includes 6 GPU servers with a total of 1TB of main memory and 8 GPU TESLA V100 processors per server, related networking switches, cables and power supplies. The proposed instruments will enable novel challenging ongoing and future research at CBIM on robust machine learning to data variability, coupling of machine learning and physics-based modeling, explainable and interpretable machine learning and AI for scene understanding and generalizability to unseen objects and scene classes, real time data analytics, dynamic event recognition, human-inspired computations, human behavior analytics and natural language understanding. This basic research requires the use of large amounts of data and is computationally intensive and CBIM currently does not have this level of computational resources. The requested instrumentation will support several existing and planned new projects at CBIM in machine learning and their applications and will enable CBIM researchers to produce state of the art research and educated graduate and undergraduate students. CBIM projects include sponsorship from ARO, AFOSR, DARPA, NSF, NIH and ONR. The requested instrumentation will allow us to enhance the current and move towards the next level of research, applications and educational activities in AI, Modeling and Machine Learning at CBIM. The proposed equipment will be used by the current 65 PhD, 15 MSc and 10 Undergraduate students at CBIM supervised by 11 faculty to improve their research and education in machine learning and related applications through the use of the best GPU systems. The requested instrumentation will be installed, administered and maintained for CBIM through systems experts from the Laboratory for Computer Science at Rutgers University.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 09, 2020
- Source ID
- W911NF2010046
Entities
People
- Dimitri Metaxas
Organizations
- Army Contracting Command
- Rutgers University
- United States Army