This $286,991 Project Grant, awarded by the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program (CFDA 47.070), will fund collaborative research on "approximate coded computing" for large-scale distributed machine learning and artificial intelligence applications. The research aims to: (i) accelerate computation by overcoming system bottlenecks, (ii) ensure accurate computation in the presence of hardware errors, and (iii) enable data-processing approaches that adhere to privacy constraints. The project will develop new coding schemes, analyses, and techniques that leverage redundancy from coding theory to address these challenges. The work will be conducted in three closely connected areas: fault-tolerant approximate matrix multiplication, fault-tolerant approximate nonlinear computations, and differentially private computations. The project will run from October 1, 2024 to May 31, 2026 and is led by the Georgia Tech Research Corporation.
Generated 1/28/25, 8:15 AM