Learning Hierarchical Features with Joint Latent Space Energy-Based Prior
Authors:
- 1 October 2023
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2218-2227
Abstract
This paper studies the fundamental problem of multilayer generator models in learning hierarchical representations. The multi-layer generator model that consists of multiple layers of latent variables organized in a top-down architecture tends to learn multiple levels of data abstraction. However, such multi-layer latent variables are typically parameterized to be Gaussian, which can be less informative in capturing complex abstractions, resulting in limited success in hierarchical representation learning. On the other hand, the energy-based (EBM) prior is known to be expressive in capturing the data regularities, but it often lacks the hierarchical structure to capture different levels of hierarchical representations. In this paper, we propose a joint latent space EBM prior model with multi-layer latent variables for effective hierarchical representation learning. We develop a variational joint learning scheme that seamlessly integrates an inference model for efficient inference. Our experiments demonstrate that the proposed joint EBM prior is effective and expressive in capturing hierarchical representations and modelling data distribution.Keywords
All Related Versions
This publication has 9 references indexed in Scilit:
- Learning Joint Latent Space EBM Prior Model for Multi-layer Generator2023
- Learning Multi-layer Latent Variable Model via Variational Optimization of Short Run MCMC for Approximate Inference2020
- On the Anatomy of MCMC-Based Maximum Likelihood Learning of Energy-Based ModelsProceedings of the AAAI Conference on Artificial Intelligence, 2020
- MoCoGAN: Decomposing Motion and Content for Video Generation2018
- Long Text Generation via Adversarial Training with Leaked InformationProceedings of the AAAI Conference on Artificial Intelligence, 2018
- Temporal Generative Adversarial Nets with Singular Value Clipping2017
- Alternating Back-Propagation for Generator NetworkProceedings of the AAAI Conference on Artificial Intelligence, 2017
- SeqGAN: Sequence Generative Adversarial Nets with Policy GradientProceedings of the AAAI Conference on Artificial Intelligence, 2017
- MCMC Using Hamiltonian Dynamics2011