Metrics for Deep Generative Models

Abstract

Neural samplers such as variational autoencoders (VAEs) or generative adversarial networks (GANs) approximate distributions by transforming samples from a simple random source---the latent space---to samples from a more complex distribution represented by a dataset. While the manifold hypothesis implies that the density induced by a dataset contains large regions of low density, the training criterions of VAEs and GANs will make the latent space densely covered. Consequently points that are separated by low-density regions in observation space will be pushed together in latent space, making stationary distances poor proxies for similarity. We transfer ideas from Riemannian geometry to this setting, letting the distance between two points be the shortest path on a Riemannian manifold induced by the transformation. The method yields a principled distance measure, provides a tool for visual inspection of deep generative models, and an alternative to linear interpolation in latent space. In addition, it can be applied for robot movement generalization using previously learned skills. The method is evaluated on a synthetic dataset with known ground truth; on a simulated robot arm dataset; on human motion capture data; and on a generative model of handwritten digits.

Published on the 21st International Conference on Artificial Intelligence and Statistics (AISTATS), 2018

References in corpus (2)

• Metrics for Deep Generative Models
• Reinterpreting Importance-Weighted Autoencoders

Cited by in corpus (18)

• Data Augmentation in High Dimensional Low Sample Size Setting Using a Geometry-Based Variational Autoencoder
• Deep generative models in inversion: a review and development of a new approach based on a variational autoencoder
• Interpreting the Latent Space of GANs for Semantic Face Editing
• Gradient-based deterministic inversion of geophysical data with Generative Adversarial Networks: is it feasible?
• Metrics for Deep Generative Models
• Change Detection in Graph Streams by Learning Graph Embeddings on Constant-Curvature Manifolds
• Amortized Variational Inference: A Systematic Review
• DeepFlow: History Matching in the Space of Deep Generative Models
• Only Bayes should learn a manifold (on the estimation of differential geometric structure from data)
• Geometry-Aware Hamiltonian Variational Auto-Encoder
• Pythae: Unifying Generative Autoencoders in Python -- A Benchmarking Use Case
• Continuous Conditional Generative Adversarial Networks (cGAN) with Generator Regularization
• Learning Flat Latent Manifolds with VAEs
• Variational Autoencoder with Learned Latent Structure
• Uniform Interpolation Constrained Geodesic Learning on Data Manifold
• Fast Approximate Geodesics for Deep Generative Models
• Pulling back information geometry
• Learning Identity-Preserving Transformations on Data Manifolds