Data-Dependent Conditional Priors for Unsupervised Learning of Multimodal Data
Autor: | Magda Gregorova, Alexandros Kalousis, Frantzeska Lavda |
---|---|
Rok vydání: | 2020 |
Předmět: |
0209 industrial biotechnology
generative models Computer science Gaussian General Physics and Astronomy lcsh:Astrophysics 02 engineering and technology Latent variable Article symbols.namesake 020901 industrial engineering & automation Joint probability distribution lcsh:QB460-466 Prior probability 0202 electrical engineering electronic engineering information engineering lcsh:Science Independence (probability theory) business.industry Sampling (statistics) Pattern recognition lcsh:QC1-999 VAE symbols Unsupervised learning lcsh:Q 020201 artificial intelligence & image processing Artificial intelligence learned prior business lcsh:Physics MNIST database |
Zdroj: | Entropy Volume 22 Issue 8 Entropy, Vol 22, Iss 888, p 888 (2020) |
ISSN: | 1099-4300 |
DOI: | 10.3390/e22080888 |
Popis: | One of the major shortcomings of variational autoencoders is the inability to produce generations from the individual modalities of data originating from mixture distributions. This is primarily due to the use of a simple isotropic Gaussian as the prior for the latent code in the ancestral sampling procedure for data generations. In this paper, we propose a novel formulation of variational autoencoders, conditional prior VAE (CP-VAE), with a two-level generative process for the observed data where continuous z and a discrete c variables are introduced in addition to the observed variables x. By learning data-dependent conditional priors, the new variational objective naturally encourages a better match between the posterior and prior conditionals, and the learning of the latent categories encoding the major source of variation of the original data in an unsupervised manner. Through sampling continuous latent code from the data-dependent conditional priors, we are able to generate new samples from the individual mixture components corresponding, to the multimodal structure over the original data. Moreover, we unify and analyse our objective under different independence assumptions for the joint distribution of the continuous and discrete latent variables. We provide an empirical evaluation on one synthetic dataset and three image datasets, FashionMNIST, MNIST, and Omniglot, illustrating the generative performance of our new model comparing to multiple baselines. |
Databáze: | OpenAIRE |
Externí odkaz: |