2013
NeurIPS
NeurIPS 2013
RNADE: The real-valued neural autoregressive density-estimator
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.
📈
Trend Setter
- Density Estimation
🐣
Hot Topic Early Bird
- distributed representation
Authors
Topics
Machine Learning > Core Methods > Classification
Machine Learning > Core Methods > Regression
Machine Learning > Learning Types > Unsupervised Learning
Deep Learning > Architectures > Neural Networks
Deep Learning > Models > Generative Models
Deep Learning > Learning Types > Representation Learning
Machine Learning > Core Methods > Density Estimation
Keywords
unsupervised learning
density estimation
neural autoregressive model
neural autoregressive density
joint density estimation
real-valued data
tractable likelihood
generative model
mixture model
autoregressive model
distributed representation
mixture density network
neural network
neural autoregressive density estimator