Poincaré Wasserstein Autoencoder
arXiv:1901.01427
Abstract
This work presents a reformulation of the recently proposed Wasserstein autoencoder framework on a non-Euclidean manifold, the Poincaré ball model of the hyperbolic space. By assuming the latent space to be hyperbolic, we can use its intrinsic hierarchy to impose structure on the learned latent space representations. We demonstrate the model in the visual domain to analyze some of its properties and show competitive results on a graph link prediction task.
References in corpus (9)
- Semi-Supervised Classification with Graph Convolutional Networks
- Neural Discrete Representation Learning
- Hyperbolic Geometry of Complex Networks
- Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoders
- Wasserstein Auto-Encoders
- Hyperspherical Variational Auto-Encoders
- A Wrapped Normal Distribution on Hyperbolic Space for Gradient-Based Learning
- Hyperbolic Neural Networks
- On the Latent Space of Wasserstein Auto-Encoders
Cited by in corpus (15)
- Normalizing Flows: An Introduction and Review of Current Methods
- Network Geometry
- Data Augmentation in High Dimensional Low Sample Size Setting Using a Geometry-Based Variational Autoencoder
- A Wrapped Normal Distribution on Hyperbolic Space for Gradient-Based Learning
- Continuous Hierarchical Representations with Poincaré Variational Auto-Encoders
- Constant Curvature Graph Convolutional Networks
- Hyperbolic Deep Neural Networks: A Survey
- Riemannian Gaussian distributions, random matrix ensembles and diffusion kernels
- Latent Variable Modelling with Hyperbolic Normalizing Flows
- From Node Embedding To Community Embedding : A Hyperbolic Approach
- APo-VAE: Text Generation in Hyperbolic Space
- Hyperbolic Neural Networks++
- Hyperbolic Graph Embedding with Enhanced Semi-Implicit Variational Inference
- Exponential-wrapped distributions on symmetric spaces
- A Wasserstein Minimum Velocity Approach to Learning Unnormalized Models