Learning Non-Convergent Non-Persistent Short-Run MCMC Toward Energy-Based Model

Abstract

This paper studies a curious phenomenon in learning energy-based model (EBM) using MCMC. In each learning iteration, we generate synthesized examples by running a non-convergent, non-mixing, and non-persistent short-run MCMC toward the current model, always starting from the same initial distribution such as uniform noise distribution, and always running a fixed number of MCMC steps. After generating synthesized examples, we then update the model parameters according to the maximum likelihood learning gradient, as if the synthesized examples are fair samples from the current model. We treat this non-convergent short-run MCMC as a learned generator model or a flow model. We provide arguments for treating the learned non-convergent short-run MCMC as a valid model. We show that the learned short-run MCMC is capable of generating realistic images. More interestingly, unlike traditional EBM or MCMC, the learned short-run MCMC is capable of reconstructing observed images and interpolating between images, like generator or flow models. The code can be found in the Appendix.

Cited by in corpus (22)

• Denoising Diffusion Probabilistic Models
• Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling
• Improved Contrastive Divergence Training of Energy Based Models
• Hybrid Discriminative-Generative Training via Contrastive Learning
• Training Deep Energy-Based Models with f-Divergence Minimization
• Cooperative Training of Fast Thinking Initializer and Slow Thinking Solver for Conditional Learning
• Trajectory Prediction with Latent Belief Energy-Based Model
• Learning Equivariant Energy Based Models with Equivariant Stein Variational Gradient Descent
• Kernel Stein Generative Modeling
• Learning Latent Space Energy-Based Prior Model
• Variational Inference with Holder Bounds
• Unpaired Image-to-Image Translation via Latent Energy Transport
• Semi-supervised Learning by Latent Space Energy-Based Model of Symbol-Vector Coupling
• Energy-Inspired Models: Learning with Sampler-Induced Distributions
• Learning Cycle-Consistent Cooperative Networks via Alternating MCMC Teaching for Unsupervised Cross-Domain Translation
• Learning Discrete Energy-based Models via Auxiliary-variable Local Exploration
• Variational (Gradient) Estimate of the Score Function in Energy-based Latent Variable Models
• Learning Deep Representation with Energy-Based Self-Expressiveness for Subspace Clustering
• Generative Text Modeling through Short Run Inference
• RNA Alternative Splicing Prediction with Discrete Compositional Energy Network
• Hamiltonian Dynamics with Non-Newtonian Momentum for Rapid Sampling
• Learning Proposals for Practical Energy-Based Regression