Fri 21 February 2020
Quasiorthonormal Encoding
Posted by Haw-minn Lu in Machine Learning
Introduction
Quasiorthonormal encoding (QOE) is a generalization of the one-hot encoding and exploits the fact that in high dimensional vector spaces, two random vectors are almost always orthogonal. The concept originated with Kůrková and Kainen almost 25 years ago. The principle is the same as one-hot encoding when you view …