Learning Deep Latent Gaussian Models with Markov Chain Monte Carlo

By Matthew D. Hoffman. 2017

94 citations recorded by the source corpus.

Subfields: Generative Models, 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 Generative Models

In Probabilistic & Bayesian Methods

Citations in the corpus

Stochastic Normalizing Flows (2020, background): “It has been demonstrated that combining learnable proposals/transformations with stochastic sampling techniques can improve expressiveness of the proposals (Salimans et al., 2015; Levy et al., 2017; Song et al., 2017; Hoffman et al., 2019; Hoffman, 2017).”
Is Generator Conditioning Causally Related to GAN Performance? (2018, methodology): “For completeness, we also compute – following Hoffman (2017), who does the same for variational autoencoders – the spectrum of the average (across a batch of z) Jacobian.”
NeuTra-lizing Bad Geometry in Hamiltonian Monte Carlo Using Neural Transport (2019, methodology): “To incorporate NeuTra HMC into these models we build upon the interleaved training procedure of Hoffman (2017). The parameters of the approximate posterior and the transport map are trained using the standard ELBO.”
Hamiltonian Variational Auto-Encoder (2018, methodology): “Finally, in Hoffman [10], an alternative approach is used where the gradient of the true likelihood,∇θL(θ;x), is directly approximated by using Fisher’s identity and HMC to obtain approximate samples from pθ(z|x).”
Amortized Inference Regularization (2018, background): “We believe that variational family expansion by Monte Carlo methods [26] may exhibit the same attenuating effect on AIR and recommend this as an additional research direction.”
Kernel Implicit Variational Inference (2017, methodology): “This problem has been pointed out previously, both in VAE works (Sønderby et al., 2016; Hoffman, 2017) and BNN works (Blundell et al., 2015), and can be explained by the looseness of the variational bound when factorized approximation is used (Trippe & Turner, 2017).”
Avoiding Latent Variable Collapse With Generative Skip Models (2018, methodology): “We follow Hoffman and Johnson [2016] and approximate this mutual information using Monte Carlo estimates of the two KL terms.”
Generative Models of Visually Grounded Imagination (2017, methodology): “Hoffman (2017) uses MCMC to fit a latent Gaussian model, which can in principle handle missing data; however, he initializes the Markov chain with the posterior mode computed by an inference network, which cannot easily handle missing inputs.”
Undirected Graphical Models as Approximate Posteriors (2019, methodology): “Variational Inference and MCMC: A common alternative to variational inference is to use MCMC to sample directly from true posteriors in generative models (Welling & Teh, 2011; Salimans et al., 2015; Wolf et al., 2016; Hoffman, 2017; Li et al., 2017; Caterini et al., 2018).”