论文标题

高维计算的理论观点

A Theoretical Perspective on Hyperdimensional Computing

论文作者

Thomas, Anthony, Dasgupta, Sanjoy, Rosing, Tajana

论文摘要

高维(HD)计算是一组神经启发的方法,用于获得数据的高维,低精度,分布式数据。这些表示可以与简单的,神经合理的算法结合使用,以实现各种信息处理任务。高清计算最近引起了计算机硬件社区的浓厚兴趣,作为解决学习问题的节能,低延迟和噪音的工具。在这篇综述中,我们提出了对高清计算的理论基础的统一处理,重点是对学习的适用性。

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.

扫码加入交流群

加入微信交流群

微信交流群二维码

扫码加入学术交流群,获取更多资源