Building the Paradigm of Big Model + Big Computational Platform for Universal Representation Learning and Deep Consensus Learning
Abstract
In this proposed effort, we plan on building a powerful and flexible GPU computational infrastructure platform. The requested equipment consists of two high-end GPU servers each of which is configured with 10 Nvidia A100 cards. The platform will be used by PI Wu s laboratory of interpretable Visual Modeling and Computing (iVMCL) in the department of Electrical and Computer Engineering at NC State University (NCSU). The proposed computational platform will enable PI Wu s lab to address ongoing research projects on learning big universal representation models for computer vision and natural language processing applications, as well as to smoothly prepare for future ones. This new platform will complement and match the input modalities already present at the iVMCL lab, and expand current capabilities in aggregating, parsing, fusing and ultimately analyzing and understanding heterogeneous Big Data in the following applications relevant to the Department of Defense (DoD): (i) Developing a Grammar-Guided Universal Representation Learning Framework for analyzing and understanding images, videos, audio, point-cloud and text in a unified modeling framework. (ii) Developing a Deep Cooperative and Compositional Learning-to-Learn Framework for learning and adapting grammar-guided deep learning models using less labeled data. (iii) Developing a Deep Consensus Learning Framework for joint reconfigurable structured input synthesis (such as layout-to-image or video synthesis) and weakly-supervised structured output prediction (such as image or video parsing). Not only will the proposed GPU cluster enable PI Wu s lab to investigate and pursue those potentially high-impact research topics, but also be beneficial to students (undergraduates and graduates) in the classes regularly taught by PI Wu, as well as other research groups at NCSU since the proposed cluster will be seamlessly integrated, and bring new feature, into the university s computing platform.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 03, 2022
- Source ID
- W911NF2210010
Entities
People
- Tianfu Wu
Organizations
- Army Contracting Command
- North Carolina State University
- United States Army