Learning Latent Space Energy-Based Prior Model
This work addresses the challenge of improving generative modeling and anomaly detection for researchers and practitioners in machine learning, though it appears incremental as it builds on existing EBM and generator frameworks.
The authors tackled the problem of learning energy-based models (EBMs) by proposing to train them in the latent space of a generator model, enabling joint learning through maximum likelihood with efficient MCMC sampling. The result is a model that shows strong performance in image and text generation and anomaly detection, as evidenced by the claim of effective capture of data regularities.
We propose to learn energy-based model (EBM) in the latent space of a generator model, so that the EBM serves as a prior model that stands on the top-down network of the generator model. Both the latent space EBM and the top-down network can be learned jointly by maximum likelihood, which involves short-run MCMC sampling from both the prior and posterior distributions of the latent vector. Due to the low dimensionality of the latent space and the expressiveness of the top-down network, a simple EBM in latent space can capture regularities in the data effectively, and MCMC sampling in latent space is efficient and mixes well. We show that the learned model exhibits strong performances in terms of image and text generation and anomaly detection. The one-page code can be found in supplementary materials.