Abstract:Hyperdimensional (HD) computing is a set of neurally inspired methods for obtaining high-dimensional, low-precision, distributed representations of data. These representations can be combined with simple, neurally plausible algorithms to effect a variety of information processing tasks. HD computing has recently garnered significant interest from the computer hardware community as an energy-efficient, low-latency, and noise-robust tool for solving learning problems. In this review, we present a unified treatment of the theoretical foundations of HD computing with a focus on the suitability of representations for learning.

What problem does this paper attempt to address?

The problems that this paper attempts to solve are: How to theoretically understand and analyze the representation methods of hyperdimensional (HD) computing and their applicability in learning tasks. Specifically, the author hopes to answer the following key questions: 1. **How can single items, item sets, and item sequences be reliably represented and stored in H - space?** 2. **What types of noise can H - space tolerate?** 3. **Which structures in the input x - space are preserved when mapped to H - space?** 4. **How expressive are linear classifiers on φ - representations?** ### Background and Motivation Hyperdimensional computing is a neuroscience - inspired method for generating high - dimensional, low - precision, distributed data representations. These representations can be combined with simple neural algorithms to perform various information - processing tasks. In recent years, HD computing has received extensive attention from the computer hardware community due to its high efficiency, low latency, and strong anti - noise ability. However, although HD computing performs well in practical applications, relatively little research has been done on its theoretical basis. This paper aims to fill this gap by providing a unified theoretical framework to understand and analyze HD computing models and exploring their applicability and performance in general learning tasks. ### Main Objectives 1. **Introduce hyperdimensional computing**: Introduce the basic concepts and principles of hyperdimensional computing to readers in the field of machine learning. 2. **Establish a mathematical framework**: Develop a specific mathematical framework to formalize and extensively study the properties of HD computing models. Through answering these questions, the author hopes to provide a solid theoretical foundation for the application and development of HD computing and promote its application in more fields.

A Theoretical Perspective on Hyperdimensional Computing

Classification using Hyperdimensional Computing: A Review

Robust In-Memory Computing with Hyperdimensional Stochastic Representation

Hyperdimensional computing: A fast, robust, and interpretable paradigm for biological data

In-memory hyperdimensional computing

Hyperdimensional computing: a fast, robust and interpretable paradigm for biological data

RegHD: Robust and Efficient Regression in Hyper-Dimensional Learning System

Generalized Holographic Reduced Representations

Task-Projected Hyperdimensional Computing for Multi-Task Learning

A Survey on Hyperdimensional Computing aka Vector Symbolic Architectures, Part I: Models and Data Transformations

Laplace-HDC: Understanding the geometry of binary hyperdimensional computing

Efficient Hyperdimensional Computing

Hyperdimensional Computing vs. Neural Networks: Comparing Architecture and Learning Process

SearcHD: A Memory-Centric Hyperdimensional Computing with Stochastic Training

Efficient Hyperdimensional Computing With Spiking Phasors

uHD: Unary Processing for Lightweight and Dynamic Hyperdimensional Computing

Hyperdimensional computing with holographic and adaptive encoder

Computing with Residue Numbers in High-Dimensional Representation

Linear Codes for Hyperdimensional Computing

A Survey on Hyperdimensional Computing aka Vector Symbolic Architectures, Part II: Applications, Cognitive Models, and Challenges

Energy Efficient In-memory Hyperdimensional Encoding for Spatio-temporal Signal Processing