Variational Rejection Sampling
By Aditya Grover, Ramki Gummadi, Miguel L'azaro-Gredilla, Dale Schuurmans, Stefano Ermon. 2018
37 citations recorded by the source corpus.
Learning latent variable models with stochastic variational inference is challenging when the approximate posterior is far from the true posterior, due to high variance in the gradient estimates. We propose a novel rejection sampling step that discards samples from the variational posterior which are assigned low likelihoods by the model. Our approach provides an arbitrarily accurate approximation of the true posterior at the expense of extra computation. Using a new gradient estimator for the resulting unnormalized proposal distribution, we achieve average improvements of 3.71 nats and 0.21 nats over state-of-the-art single-sample and multi-sample alternatives respectively for estimating marginal log-likelihoods using sigmoid belief networks on the MNIST dataset.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Probabilistic & Bayesian Methods
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
- Yoshua Bengio
- Stefano Ermon
- Sergey Levine
- Quoc V. Le
- Ruslan Salakhutdinov
- Mohit Bansal
- Kyunghyun Cho
- Pieter Abbeel
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Trevor Darrell
- Stefano Ermon
- Dacheng Tao
- Ming-Hsuan Yang
- Jitendra Malik
- Xiaogang Wang
- Yoshua Bengio
In Probabilistic & Bayesian Methods
- Stefano Ermon
- Yoshua Bengio
- Jiaming Song
- Jascha Narain Sohl-Dickstein
- Andriy Mnih
- Aditya Grover
- Laurent Dinh
- Hugo Larochelle
Citations in the corpus
- Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling (2020, background): “[24] considered rejection sampling in latent space in encoder-decoder models.”
- Discriminator Rejection Sampling (2018, methodology): “An independent rejection sampling approach was proposed by Grover et al. (2018) in the latent space of variational autoencoders for improving samples from the variational posterior.”
- Resampled Priors for Variational Autoencoders (2018, methodology): “We evaluate LARS on a suite of standard datasets (static and dynamically binarized MNIST [9], Omniglot [8], and FashionMNIST [16]) using several architectures: (i) a VAE with single latent layer”
- Bias Correction of Learned Generative Models using Likelihood-Free Importance Weighting (2019, methodology): “Finally, we note resampling based techniques have been extensively studied in the context of improving variational approximations for latent variable generative models [49–52].”
- A RAD approach to deep mixture models (2019, background): “It provides a more constrainted but less sample wasteful approach than rejection sampling (Grover et al., 2018; Azadi et al., 2019; Bauer and Mnih, 2019) of taking into account the untransformed sample z before selecting the mixture component k.”
- An approximate sampler for energy-based models with divergence diagnostics (2022, unknown): “Variational rejection sampling (Grover et al., 2018) uses a relaxation of Eq.”
- Refining Deep Generative Models via Discriminator Gradient Flow (2020, methodology): “Other Related Work Monte Carlo techniques have been used for improving various components in generative models, e.g., Grover et al. (2018) proposed Variational Rejection Sampling which performs rejection sampling in the latent space of VAEs to improve the variational posterior and Grover et al.…”