Unsupervised Variational Bayesian Learning of Nonlinear Models

Abstract

In this paper we present a framework for using multi-layer per- ceptron (MLP) networks in nonlinear generative models trained by variational Bayesian learning. The nonlinearity is handled by linearizing it using a Gauss–Hermite quadrature at the hidden neu- rons. This yields an accurate approximation for cases of large pos- terior variance. The method can be used to derive nonlinear coun- terparts for linear algorithms such as factor analysis, independent component/factor analysis and state-space models. This is demon- strated with a nonlinear factor analysis experiment in which even 20 sources can be estimated from a real world speech data set.

Cite

Text

Honkela and Valpola. "Unsupervised Variational Bayesian Learning of Nonlinear Models." Neural Information Processing Systems, 2004.

Markdown

[Honkela and Valpola. "Unsupervised Variational Bayesian Learning of Nonlinear Models." Neural Information Processing Systems, 2004.](https://mlanthology.org/neurips/2004/honkela2004neurips-unsupervised/)

BibTeX

@inproceedings{honkela2004neurips-unsupervised,
  title     = {{Unsupervised Variational Bayesian Learning of Nonlinear Models}},
  author    = {Honkela, Antti and Valpola, Harri},
  booktitle = {Neural Information Processing Systems},
  year      = {2004},
  pages     = {593-600},
  url       = {https://mlanthology.org/neurips/2004/honkela2004neurips-unsupervised/}
}