Learning a Continuous Hidden Variable Model for Binary Data

A directed generative model for binary data using a small number of hidden continuous units is investigated. A clipping nonlinearity distinguishes the model from conventional principal components analysis. The relationships between the correlations of the underlying continuous Gaussian variables and the binary output variables are utilized to learn the appropriate weights of the network. The advantages of this approach are illustrated on a translationally invariant binary distribution and on handwritten digit images.

• Type: Conference
• Year: 1998
• Where: NIPS
• Authors: Daniel D. Lee, Haim Sompolinsky
• Keywords: Binary Output Variables, Continuous Gaussian Variables, Directed Generative Model, NIPS 1998, NIPS 2007