loading page

KL-Autoencoders for manifold learning
  • Khoa Doan
Khoa Doan
Virginia Tech
Author Profile

Abstract

We propose efficient nonlinear dimensionality reduction method based on information theory and autoencoders. We extends autoencoders to learn also the manifold of the original data. The methods are demonstrated on important category of earth science dataset that is obtained from the popular afternoon constellation system.