* 18 pages. We propose a novel invertible dimension reduction process
for manifold learning with a neural network implementation, and explore the
inherent difficulty of manifold learning in real-world secnarios by the way Access Paper or Ask Questions
