Maximum Entropy Generators for Energy-Based Models
This work addresses the intractable log-likelihood gradient problem in energy-based models for researchers in generative modeling, offering a stable and efficient alternative to GANs with broad applications in image generation and anomaly detection.
The authors tackled the challenge of maximum likelihood estimation in energy-based models by proposing a method that learns both the energy function and an amortized approximate sampling mechanism using a neural generator network, achieving competitive Inception and FID scores for image generation and state-of-the-art performance in anomaly detection without mode collapse.
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.