IEEE Circuits and Systems Magazine - Q2 2020 - 30

Feature

Classification Using
Hyperdimensional
Computing: A Review
Lulu Ge, Student Member, IEEE, and Keshab K. Parhi, Fellow, IEEE

Abstract
Hyperdimensional (HD) computing is built upon its unique data
type referred to as hypervectors. The dimension of these hypervectors is typically in the range of tens of thousands. Proposed to
solve cognitive tasks, HD computing aims at calculating similarity
among its data. Data transformation is realized by three operations,
including addition, multiplication and permutation. Its ultra-wide
data representation introduces redundancy against noise. Since
information is evenly distributed over every bit of the hypervectors,
HD computing is inherently robust. Additionally, due to the nature
of those three operations, HD computing leads to fast learning ability, high energy efficiency and acceptable accuracy in learning and
classification tasks. This paper introduces the background of HD
computing, and reviews the data representation, data transformation, and similarity measurement. The orthogonality in high dimensions presents opportunities for flexible computing. To balance the
tradeoff between accuracy and efficiency, strategies include but
are not limited to encoding, retraining, binarization and hardware
acceleration. Evaluations indicate that HD computing shows great
potential in addressing problems using data in the form of letters,
signals and images. HD computing especially shows significant
promise to replace machine learning algorithms as a light-weight
classifier in the field of internet of things (IoTs).
Digital Object Identifier 10.1109/MCAS.2020.2988388
Date of current version: 3 June 2020

IEEE CIRCUITS AND SYSTEMS MAGAZINE

©ISTOCKPHOTO/ MONSITJ

I. Introduction
he emergence of hyperdimensional (HD) computing is based on the cognitive model developed by
Kanerva [1]. HD computing grew out of cognitive
science in answer to the binding problem of connectionist (neural-net) models. When variables and their values
are superposed over the same vector, representing which
value is associated with which variable requires a formal
model. This was initially solved using tensor product variable binding by Smolensky [2] and later by Plate [3] using
holographic reduced representation (HRR). The advantage of HRR over tensor product is that it keeps vector dimensionality constant. Systems based on these representations go by many names: HRR, HD, binary spatter code
(BSD) [4], binary sparse distributed code (BSDC) [5], multiply-add-permute (MAP) [6], vector symbolic architecture (VSA) [7], and semantic pointer architecture. All rely
on high dimensionality, randomness, abundance of nearly
orthogonal vectors and computing in superposition.
Instead of computing traditional numerical values, HD
computing performs cognition tasks-such as face detection, language classification, speech recognition, i-mage

SECOND QUARTER 2020

IEEE Circuits and Systems Magazine - Q2 2020

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q2 2020