Restricted Research - Award List, Note/Discussion Page
Fiscal Year: 2023
2418 The University of Texas at San Antonio (144306)
Principal Investigator: John, Eugene
Total Amount of Contract, Award, or Gift (Annual before 2011): $ 100,002
Exceeds $250,000 (Is it flagged?): No
Start and End Dates: 1/1/23 - 12/31/25
Restricted Research: YES
Academic Discipline: Engineering and Integrated Design
Department, Center, School, or Institute: Ctr Excel Engr Research & Educ
Title of Contract, Award, or Gift: Ultra Low-Energy Ultra Low-Latency Machine Learning using Weightless Neural Networks
Name of Granting or Contracting Agency/Entity:
University of Texas at Austin
CFDA: 0
Program Title: none
Note:
SAMs 1.1.1; The objectives of this research project are to design Weightless Neural Networks (WNNs) with low energy and high accuracy. Artificial Intelligence Foundation models will also be investigated. Verilog and high level synthesis models will be developed. We will explore and adapt the aforementioned ideas to create ultra-low energy low-latency WNNs as well as to improve their accuracy. For this research, we will use CIFAR-10, SVHN and other complex image data sets. It is not clear that ImageNet can be handled, but WNNs are highly suited for applications such as arrhythmia classification and regression [9], and hence developing WNNs for such applications will also be pursued. Simulation models as well as FPGA prototypes will be developed. Area, energy, accuracy, latency and memory consumption will be compared on FPGA models and ASIC models. Feasibility of implementing lookup tables using TCAMs will be investigated. GPU models will also be used as a baseline for comparison.
Discussion: No discussion notes