Hyper-Dimensional Computing and its Applications in tinyML

Document Type

Article

Authors

Ellis A. Weglewski, University of Minnesota MorrisFollow

Abstract

As computing systems enter the realm of nano form levels, new fields of computational development have spawned, each posing their own set of challenges. Amongst these fields is Tiny Machine Learning (tinyML), which aims to install machine learning on tiny embedded systems. The restrictions imposed upon algorithms by the limited hardware of nano-scale tiny systems make contemporary approaches to machine learning non-contenders. Hyperdimensional computing is an approach to representing data as high-dimensional vectors which allows for one-pass encoding and quick all-encompassing comparison operations via an associative memory. This approach is power-efficient, robust, and can be done in-memory, all of which make it a viable candidate for tinyML.

Recommended Citation

Weglewski, Ellis A. () "Hyper-Dimensional Computing and its Applications in tinyML," Scholarly Horizons: University of Minnesota, Morris Undergraduate Journal: Vol. 11: Iss. 2, Article 14.
DOI: https://doi.org/10.61366/2576-2176.1147
Available at: https://digitalcommons.morris.umn.edu/horizons/vol11/iss2/14

Primo Type

Article