Effect of Sparsity on Disentanglement: Several sparsity-inducing penalties are implemented within the Variational Autoencoders framework to quantify their effect on learning disentangled representations (a more detailed description can be found on my Research School page).
Disentangling via Transformation-Robust Representations: Invariance against groups of transformations on particular latent space dimensions is imposed, allowing to incorporate prior knowledge on a weakly-supervised level.
Contact
Chair Representative:
Prof. Dr. Christoph Lippert Professor for Digital Health & Machine Learning Room: G-2.1.23 Tel.: +49-(0)331 5509-4850 E-Mail: office-lippert(at)hpi.de
Office:
Campus III, Haus G2 Room: G-2.1.22 Tel.: +49-(0)331 5509-4850 Fax: +49-(0)331 5509-4849 E-Mail: office-lippert(at)hpi.de
Visiting address:
Campus III Building G2 Rudolf-Breitscheid-Straße 187 14482 Potsdam, Germany