inferlo.GaussianModel

class inferlo.GaussianModel

Gaussian graphical model

A Gaussian graphical model is defined by an undirected graph G = (V;E), where V is the set of nodes and E is the set of edges, and a collection of jointly Gaussian random variables x= (x_i; i in V). The probability density is given by p(x) ∝ exp{-1/2*x^T*J*x + h^T*x} where J is a symmetric, positive definite matrix, that is sparse so as to respect the graph G: if {i,j} not in E, then J_{i,j} = 0

For more information please refer to ‘Walk-SumsandBeliefPropagationinGaussianGraphicalModels’ by Dmitry M. Malioutov, Jason K. Johnson, Alan S. Willsky : http://ssg.mit.edu/group/willsky/publ_pdfs/185_pub_MLR.pdf

__init__(J: array, h: array, domain=None)

Parameters:

• num_variables – Number of variables in the model.

• domain – Default domain of each variable.

Methods

__init__(J, h[, domain])
add_factor(factor)	Adds factor.
copy()	Makes a copy of itself.
draw_factor_graph(ax)	Draws the factor graph.
evaluate(x)	Returns value of non-normalized pdf in point.
from_model(model)	Creates copy of a given model.
get_factor_graph()	Builds factor graph for the model.
get_factors()	Returns all factors.
get_max_domain_size()	Returns the biggest domain size over all variables.
get_symbolic_variables()	Prepares variables for usage in expressions.
get_variable(idx)	Returns variable by its index.
get_variables()	Returns all variables.
infer(**kwargs)	Performs inference.
max_likelihood([algorithm])	Finds most probable state.
max_likelihood_bruteforce()	Evaluates most likely state in a very inefficient way.
part_func_bruteforce()	Evaluates partition function in very inefficient way.
sample(num_samples[, algorithm])	Generates samples.

Attributes

G
J
h
n

infer(**kwargs)

Performs inference.