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

Abstract

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.

Under review as a conference paper at ICLR 2017

References in corpus (12)

• Adam: A Method for Stochastic Optimization
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop
• Energy-based Generative Adversarial Network
• f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization
• Learning to learn by gradient descent by gradient descent
• Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm
• Learning to Optimize
• Deep Directed Generative Models with Energy-Based Probability Estimation
• Inference Networks for Sequential Monte Carlo in Graphical Models
• Integral approximation by kernel smoothing
• Hierarchical Variational Models

Cited by in corpus (39)

• Improved Training of Wasserstein GANs
• Reinforcement Learning with Deep Energy-Based Policies
• Adversarial Feature Matching for Text Generation
• Measuring Sample Quality with Kernels
• Discrete Sequential Prediction of Continuous Actions for Deep RL
• Improving the Improved Training of Wasserstein GANs: A Consistency Term and Its Dual Effect
• Gradient Estimators for Implicit Models
• GenDICE: Generalized Offline Estimation of Stationary Values
• Activation Maximization Generative Adversarial Nets
• Stacked Generative Adversarial Networks
• Stein Variational Gradient Descent With Matrix-Valued Kernels
• A Non-Asymptotic Analysis for Stein Variational Gradient Descent
• Advances in Variational Inference
• Boosting Deep Learning Risk Prediction with Generative Adversarial Networks for Electronic Health Records
• Approximate Inference with Amortised MCMC
• Riemannian Stein Variational Gradient Descent for Bayesian Inference
• Towards Robust, Locally Linear Deep Networks
• Learning Model Reparametrizations: Implicit Variational Inference by Fitting MCMC distributions
• Learning Deep Energy Models: Contrastive Divergence vs. Amortized MLE
• Stein Neural Sampler
• The reproducing Stein kernel approach for post-hoc corrected sampling
• Training Deep Energy-Based Models with f-Divergence Minimization
• Emergence of Theory of Mind Collaboration in Multiagent Systems
• Two Methods For Wild Variational Inference
• Variational Implicit Processes
• Stochastic Maximum Likelihood Optimization via Hypernetworks
• Stochastic Stein Discrepancies
• Cooperative Training of Fast Thinking Initializer and Slow Thinking Solver for Conditional Learning
• Deep Generative Learning via Variational Gradient Flow
• Score-Guided Generative Adversarial Networks
• Kernel Stein Generative Modeling
• Gradient Layer: Enhancing the Convergence of Adversarial Training for Generative Models
• Distribution Aware Active Learning
• Spherical Sliced-Wasserstein
• Scalable Approximate Inference and Some Applications
• Stabilizing Training of Generative Adversarial Nets via Langevin Stein Variational Gradient Descent
• Generative Particle Variational Inference via Estimation of Functional Gradients
• Adversarial Variational Bayes Methods for Tweedie Compound Poisson Mixed Models
• FRAME Revisited: An Interpretation View Based on Particle Evolution