JSAI2019

Presentation information

General Session

General Session » [GS] J-2 Machine learning

[2Q5-J-2] Machine learning: advanced models

Wed. Jun 5, 2019 5:20 PM - 7:00 PM Room Q (6F Meeting room, Bandaijima bldg.)

Chair:Masataro Asai Reviewer:Hikaru Kajino

6:20 PM - 6:40 PM

[2Q5-J-2-04] Variational Auto-Encoder On Stiefel Space

〇Takaaki Sanjoh1, Junpei Komiyama2, Masashi Toyoda2, Masaru Kitsuregawa2,3 (1. The University of Tokyo, 2. Institute of Industrial Science, The University of Tokyo, 3. National Institute of Informatics)

Keywords:Variational Auto-Encoder, Manifold Learning, Stiefel space

This paper presents a reformulation of Variational Auto-Encoder (VAE) framework on a non-Euclidean manifold, the Stiefel space $\stV$. By assuming the latent space to be Stiefel manifold, we can use its
intrinsic orthonormality to impose structure on the learned latent space representations.
We derive an objective function and gradient descendant method for learning VAE using a probabilistic distribution on the Stiefel space.