A Wasserstein Minimum Velocity Approach to Learning Unnormalized Models
arXiv:2002.07501
Abstract
Score matching provides an effective approach to learning flexible unnormalized models, but its scalability is limited by the need to evaluate a second-order derivative. In this paper, we present a scalable approximation to a general family of learning objectives including score matching, by observing a new connection between these objectives and Wasserstein gradient flows. We present applications with promise in learning neural density estimators on manifolds, and training implicit variational and Wasserstein auto-encoders with a manifold-valued prior.
AISTATS 2020
References in corpus (13)
- GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
- Searching for Activation Functions
- Wasserstein Auto-Encoders
- Stein Variational Gradient Descent as Gradient Flow
- Hyperspherical Variational Auto-Encoders
- Langevin diffusions and the Metropolis-adjusted Langevin algorithm
- Density Estimation in Infinite Dimensional Exponential Families
- Interpretation and Generalization of Score Matching
- Understanding and Accelerating Particle-Based Variational Inference
- Deep Energy Estimator Networks
- Poincaré Wasserstein Autoencoder
- Score matching estimators for directional distributions
- Accelerating Langevin Sampling with Birth-death