Helmholtz machine
From Scholarpedia
This article has not yet been published; it may contain inaccuracies, unapproved changes, or be unfinished.
Dr. Geoffrey E. Hinton accepted the invitation on 4 July 2008 (self-imposed deadline: 4 October 2008).
Helmholtz machine refers to a type of multilayer neural network in which top-down connections implement a generative model of the sensory input and bottom-up connections are used to perform perceptual inference by computing a variational approximation to the true posterior distribution over the hidden states.
Dr. Geoffrey E. Hinton, University of Toronto, CANADA, was invited on 18 April 2008.
