LeHDC: Learning-Based Hyperdimensional Computing Classifier
TimeThursday, July 14th4:10pm - 4:30pm PDT
Location3005, Level 3
Event Type
Research Manuscript
Emerging Models of Computation
DescriptionThanks to its tiny storage requirement and efficient execution, Hyperdimensional Computing (HDC) is emerging as a lightweight learning framework on resource-constrained hardware. Nonetheless, the existing HDC training relies on various heuristic methods, significantly limiting their inference accuracy. In this paper, we propose a new learning-enabled HDC framework, called LeHDC, to systematically improve accuracy. Concretely, LeHDC maps the existing HDC model into an equivalent Binary Neural Network architecture, and employs a corresponding principled training strategy. Our validation shows that LeHDC outperforms previous HDC training strategies and can improve on average the inference accuracy by up to 16% against the baseline HDC.