Project Grant 2217086

This $125,000 Project Grant from the National Science Foundation's Computer and Information Science and Engineering program (CFDA #47.070) supports research at the University of California, Merced on developing optimized sparse tensor network algorithms and specialized accelerator architectures. The one-year award beginning October 1, 2022 aims to address challenges in high-dimensional data computation and analytics using tensor representations by exploring memory heterogeneity-aware data arrangements, balanced sparse tensor contraction algorithms, memoization techniques, and hardware-software co-design of sparse tensor network accelerators. The research encompasses efforts across high-performance computing, algorithms, compilers, computer architecture, and performance modeling to optimize sparse tensor networks for heterogeneous systems with various accelerators and memory types. Findings will advance state-of-the-art tensor decomposition methods for modeling higher-order and sparse data at scale.