C O D E
DigitDemo v1.0: This code package contains an interactive
demo of Gibbs sampling, mean-field reconstruction and classification for restricted
Boltzmann machines trained using four different inductive principles including
stochastic maximum likelihood (SML), contrastive divergence (CD),
ratio matching (RM) and pseudo-likelihood (PL). The MNIST data set was used as training data.
See our AISTATS 2010 paper
Inductive Principles for Restricted Boltzmann Machine Learning
for details. Note that this demo uses trained models and does not
include code for learning.
GGM-Gwishart v1.0: This MATLAB code package implements
several marginal likelihood approximation methods for non-decomposable Gaussian Graphical
Models (GGMs) under a G-Wishart prior. The methods are described in our
NIPS 2009 paper.
The approximations include BIC score, Laplace approximation, diagonal Hessian Laplace approximation,
and the Monte Carlo estimator of Atay-Kayis and Massam.