Learning to Draw Samples: With Application to Amortized MLE for Generative Adversarial Learning

By Dilin Wang, Qiang Liu. 2016

122 citations recorded by the source corpus.

We propose a simple algorithm to train stochastic neural networks to draw samples from given target distributions for probabilistic inference. Our method is based on iteratively adjusting the neural network parameters so that the output changes along a Stein variational gradient that maximumly decreases the KL divergence with the target distribution. Our method works for any target distribution specified by their unnormalized density function, and can train any black-box architectures that are differentiable in terms of the parameters we want to adapt. As an application of our method, we propose an amortized MLE algorithm for training deep energy model, where a neural sampler is adaptively trained to approximate the likelihood function. Our method mimics an adversarial game between the deep energy model and the neural sampler, and obtains realistic-looking images competitive with the state-of-the-art results.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Statistical and Nonlinear Physics, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

In Statistical and Nonlinear Physics

In Generative Models

In Probabilistic & Bayesian Methods

Citations in the corpus

Improved Training of Wasserstein GANs (2017, neutral): “07 Supervised Method Score SteinGAN [25] 6.”
Reinforcement Learning with Deep Energy-Based Policies (2017, methodology): “The sampling network trained as part of SVGD can also be viewed as tking the role of an actor in an actor-critic algorithm.”
Stacked Generative Adversarial Networks (2016, neutral): “This is consistent with results in [43, 61].”
Adversarial Feature Matching for Text Generation (2017, methodology): “Recent work considered combining autoencoders or variational autoencoders (Kingma & Welling, 2014) with GAN (Zhao et al., 2017; Larsen et al., 2016; Makhzani et al., 2015; Mescheder et al., 2017; Wang & Liu, 2016).”
Discrete Sequential Prediction of Continuous Actions for Deep RL (2017, methodology): “Existing methods use stochastic neural networks [7] or construct energy models [14] sampled with Stein variational gradient descent [22, 47].”
Adversarial Fisher Vectors for Unsupervised Representation Learning (2019, methodology): “Recently, [2, 3, 4] show that training an energy-based model (EBM) with a parameterized variational distribution reduces to a similar minimax GAN game.”
Continuous-Time Flows for Efficient Inference and Density Estimation (2017, methodology): “Similar to Wang & Liu (2017), our CTF-based approach in Section 4 provides an alternative way for this problem, by simultaneously learning an explicit energy-based data distribution (estimated density) and a generator whose generated samples match the learned data distribution.”
Variational Implicit Processes (2018, methodology): “In approximate inference context, implicit distributions have also been deployed to postulate ﬂexible approximate posterior distributions [56, 58, 60, 61, 76, 87, 90].”
Composable Deep Reinforcement Learning for Robotic Manipulation (2018, methodology): “To that end, soft Q-learning uses amortized Stein variational descent (SVGD) [33], [34] to learn a stochastic neural network to approximate samples from the desired EBM [7].”
Wild Variational Approximations (2016, methodology): “Very recently [48] applied the amortized SVGD idea to learning energy-based models.”
Gradient Estimators for Implicit Models (2017, background): “Very recently, implicit distributions have also been considered as approximate posterior distributions for Bayesian inference, e.g. see [22, 47, 19, 17, 26, 13, 20, 43].”
Two Methods for Wild Variational Inference (2016, unknown): “…work: one based on “amortizing SVGD” (Wang & Liu, 2016) which trains the inference network f(η, ξ) so that its output mimic the SVGD dynamics in order to decrease the KL divergence; another based on minimizing the KSD objective (9) which does not require to evaluate q(z) thanks to its special form.”
Adversarial Learning of a Sampler Based on an Unnormalized Distribution (2019, methodology): “We also compare with the standard SVGD [Liu and Wang, 2016] and the amortized SVGD methods [Wang and Liu, 2016], in which 512 particles are used.”
Fisher GAN (2017, unknown): “05 Unsupervised Method Score SteinGan [31] 6.”