Ordering Dimensions with Nested Dropout Normalizing Flows
arXiv:2006.08777
Abstract
The latent space of normalizing flows must be of the same dimensionality as their output space. This constraint presents a problem if we want to learn low-dimensional, semantically meaningful representations. Recent work has provided compact representations by fitting flows constrained to manifolds, but hasn't defined a density off that manifold. In this work we consider flows with full support in data space, but with ordered latent variables. Like in PCA, the leading latent dimensions define a sequence of manifolds that lie close to the data. We note a trade-off between the flow likelihood and the quality of the ordering, depending on the parameterization of the flow.
References in corpus (6)
- Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
- Improving Variational Auto-Encoders using Householder Flow
- Normalizing Flows on Riemannian Manifolds
- Emerging Convolutions for Generative Normalizing Flows
- Hybrid Models with Deep and Invertible Features
- Regularized Autoencoders via Relaxed Injective Probability Flow