Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: Autoencoding topology
(Submitted on 1 Mar 2018)
Abstract: The problem of learning a manifold structure on a dataset is framed in terms of a generative model, to which we use ideas behind autoencoders (namely adversarial/Wasserstein autoencoders) to fit deep neural networks. From a machine learning perspective, the resulting structure, an atlas of a manifold, may be viewed as a combination of dimensionality reduction and "fuzzy" clustering.
Link back to: arXiv, form interface, contact.