Maximum Entropy Generators for Energy-Based Models

Abstract

Maximum likelihood estimation of energy-based models is a challenging problem due to the intractability of the log-likelihood gradient. In this work, we propose learning both the energy function and an amortized approximate sampling mechanism using a neural generator network, which provides an efficient approximation of the log-likelihood gradient. The resulting objective requires maximizing entropy of the generated samples, which we perform using recently proposed nonparametric mutual information estimators. Finally, to stabilize the resulting adversarial game, we use a zero-centered gradient penalty derived as a necessary condition from the score matching literature. The proposed technique can generate sharp images with Inception and FID scores competitive with recent GAN techniques, does not suffer from mode collapse, and is competitive with state-of-the-art anomaly detection techniques.

References in corpus (15)

• Auto-Encoding Variational Bayes
• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Spectral Normalization for Generative Adversarial Networks
• Energy-based Generative Adversarial Network
• Adversarially Learned Inference
• f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization
• Efficient GAN-Based Anomaly Detection
• A note on the evaluation of generative models
• Deep Structured Energy Based Models for Anomaly Detection
• Better Mixing via Deep Representations
• The Numerics of GANs
• On distinguishability criteria for estimating generative models
• Deep Directed Generative Models with Energy-Based Probability Estimation
• Learning Independent Features with Adversarial Nets for Non-linear ICA

Cited by in corpus (29)

• Implicit Generation and Generalization in Energy-Based Models
• How to Train Your Energy-Based Models
• Deep Generative Modelling: A Comparative Review of VAEs, GANs, Normalizing Flows, Energy-Based and Autoregressive Models
• MSTREAM: Fast Anomaly Detection in Multi-Aspect Streams
• LOGAN: Latent Optimisation for Generative Adversarial Networks
• Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling
• Prescribed Generative Adversarial Networks
• Small-GAN: Speeding Up GAN Training Using Core-sets
• Adversarial score matching and improved sampling for image generation
• Learning Implicit Priors for Motion Optimization
• O-GAN: Extremely Concise Approach for Auto-Encoding Generative Adversarial Networks
• Learning Energy-Based Models by Diffusion Recovery Likelihood
• Top-k Training of GANs: Improving GAN Performance by Throwing Away Bad Samples
• Training Deep Energy-Based Models with f-Divergence Minimization
• Synergetic Learning Systems: Concept, Architecture, and Algorithms
• Discriminator Contrastive Divergence: Semi-Amortized Generative Modeling by Exploring Energy of the Discriminator
• No MCMC for me: Amortized sampling for fast and stable training of energy-based models
• Learning Equivariant Energy Based Models with Equivariant Stein Variational Gradient Descent
• Bounds all around: training energy-based models with bidirectional bounds
• Self-Supervised GAN Compression
• Learning Latent Space Energy-Based Prior Model
• Joint Training of Variational Auto-Encoder and Latent Energy-Based Model
• Efficient Learning of Generative Models via Finite-Difference Score Matching
• Deep Consensus Learning
• DOI: Divergence-based Out-of-Distribution Indicators via Deep Generative Models
• Explaining the effects of non-convergent sampling in the training of Energy-Based Models
• Deep Out-of-Distribution Uncertainty Quantification via Weight Entropy Maximization
• Hamiltonian Dynamics with Non-Newtonian Momentum for Rapid Sampling
• Learning Proposals for Practical Energy-Based Regression