English

Flexible Prior Distributions for Deep Generative Models

Machine Learning 2018-01-09 v2 Machine Learning

Abstract

We consider the problem of training generative models with deep neural networks as generators, i.e. to map latent codes to data points. Whereas the dominant paradigm combines simple priors over codes with complex deterministic models, we argue that it might be advantageous to use more flexible code distributions. We demonstrate how these distributions can be induced directly from the data. The benefits include: more powerful generative models, better modeling of latent structure and explicit control of the degree of generalization.

Keywords

Cite

@article{arxiv.1710.11383,
  title  = {Flexible Prior Distributions for Deep Generative Models},
  author = {Yannic Kilcher and Aurelien Lucchi and Thomas Hofmann},
  journal= {arXiv preprint arXiv:1710.11383},
  year   = {2018}
}

Comments

arXiv admin note: text overlap with arXiv:1707.09241

R2 v1 2026-06-22T22:30:58.300Z