Stein Neural Sampler

By Tianyang Hu, Zixiang Chen, Hanxi Sun, Jincheng Bai, Mao Ye, Guang Cheng. 2018

37 citations recorded by the source corpus.

We propose two novel samplers to produce high-quality samples from a given (un-normalized) probability density. The sampling is achieved by transforming a reference distribution to the target distribution with neural networks, which are trained separately by minimizing two kinds of Stein Discrepancies, and hence our method is named as Stein neural sampler. Theoretical and empirical results suggest that, compared with traditional sampling schemes, our samplers share the following three advantages: 1. Being asymptotically correct; 2. Experiencing less convergence issue in practice; 3. Generating samples instantaneously.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Posterior Meta-Replay for Continual Learning (2021, background): “Stein’s identity and the related Stein discrepancy have also been investigated to develop training methods for implicit distributions [25].”
Variational (Gradient) Estimate of the Score Function in Energy-based Latent Variable Models (2020, unknown): “Speciﬁcally, we consider the Fisher divergence of the maximum form (Hu et al., 2018; Grathwohl et al., 2020b) to measure how well p θ ( v ) approximates p D ( v ) : where F is the set of functions f : R d → R d satisfying lim || v ||→∞ p D ( v ) f ( v ) = 0 and p ( (cid:15) ) is a noise…”