Massive Computing System to Understand and Track Human Cognition
Abstract
This abstract is publicly releasableThe goal of this proposal is to create a computational cluster that will enable us to develop ar,tificial intelligence techniques to improve human learning and performance, track their brain activity, and facilitate their interac,tions with artificial agents. There is a nexus of projects at Carnegie Mellon that are pushing state-of-the-art brain imaging techni,ques, computational modeling of human cognition, and interactions of large numbers of human and artificial intelligence agents. The,proposed cluster will greatly accelerate the rate of progress on these projects, and would also contribute to the training of studen,ts and researchers, many of whom do work or will work in DoD laboratories.The proposed system consists of a mix of Intel Xeon-based,and AMD EPYC-based hardware. Computational components include: 1 login node, 2 head nodes, 3 network-based storage servers (~720TB u,nformatted capacity), 20 192GB compute nodes, 8 256GB compute nodes, 5 1TB compute nodes, and 2 512GB quad-GPU compute nodes. This s,ystem will allow us to adaptively distribute computational resources over projects while providing the flexibility to accommodate un,ique research-specific computational demands. The system will enable us to process huge volumes of fMRI and EEG data at rates commen,surate with the questions we are asking and the methods we are developing. It will make it possible to run tens of thousands of simu,lations of individuals and teams at a rate that can keep pace with our development of cognitive models, commensurate with the comple,xity of human cognition.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 16, 2022
- Source ID
- N000142212345
Entities
People
- Marcel Just
Organizations
- Carnegie Mellon University
- Office of Naval Research
- United States Navy