Asymmetric locality preserving projection and its application to k-nearest neighbor method
Yoshio Iwai, Masashi Nishiyama, Hiroki Yoshimura · 2017
In recent years, many methods for data compression and structure extraction from various types of massive data using multivariate analysis have been proposed. The locality preserving projection, which uses a symmetric similarity matrix, is one of these data compression methods. However, the similarity matrix expressing the characteristic of data may often not be symmetric in real. In this sturdy, we propose an asymmetric locality preserving projection that expands the locality preserving projection from a symmetric similarity matrix method to one that uses an asymmetric similarity matrix. We also show the experimental results of its application to the k-nearest neighbor method as an example.