RNADE: The real-valued neural autoregressive density-estimator

Benigno Uria, Iain Murray, Hugo Larochelle

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We introduce RNADE, a new model for joint density estimation of real-valued vectors. Our model calculates the density of a datapoint as the product of one-dimensional conditionals modeled using mixture density networks with shared parameters. RNADE learns a distributed representation of the data, while having a tractable expression for the calculation of densities. A tractable likelihood allows direct comparison with other methods and training by standard gradient-based optimizers. We compare the performance of RNADE on several datasets of heterogeneous and perceptual data, finding it outperforms mixture models in all but one case.
Original languageEnglish
Title of host publicationAdvances in Neural Information Processing Systems 26
Pages2175-2183
Number of pages9
Publication statusPublished - 2013

Cite this