GENERIC: Highly Efficient Learning Engine on Edge using Hyperdimensional Computing
TimeThursday, July 14th4:30pm - 4:50pm PDT
Location3005, Level 3
Event Type
Research Manuscript
Emerging Models of Computation
DescriptionHyperdimensional Computing (HDC) mimics the brain’s basic principles in performing cognitive tasks by encoding the data to high-dimensional representation and employing noncomplex learning techniques on such representation. In this paper, we first propose a novel encoding that achieves high accuracy for diverse benchmarks. Thereafter, we leverage our generic HDC encoding and propose a highly efficient ASIC accelerator suited for edge domain. Our design is the first HDC ASIC that supports both classification (train and inference) and clustering for unsupervised learning on edge. The proposed architecture is flexible in terms of main HDC parameters and can implement various applications.