inference.MCMC_Engine.MCMC_Engine Class Reference

Inheritance diagram for inference.MCMC_Engine.MCMC_Engine:

Collaboration diagram for inference.MCMC_Engine.MCMC_Engine:

Public Member Functions
def	__init__ (self, bnet, do_print=False, is_quantum=False)
def	get_unipot_list (self, node_list, num_cycles, warmup)
def	sample_node_given_markov_blanket (self, focus_node, annotated_story)
Public Member Functions inherited from inference.InferenceEngine.InferenceEngine
def	__init__ (self, bnet, do_print=False, is_quantum=False)

Public Attributes
	bnet_ord_nodes
Public Attributes inherited from inference.InferenceEngine.InferenceEngine
	bnet
	do_print
	is_quantum

Additional Inherited Members
Static Public Member Functions inherited from inference.InferenceEngine.InferenceEngine
def	print_annotated_story (annotated_story)

Detailed Description

An MCMC_Engine is an InferenceEngine that uses MCMC = Markov Chain Monte
Carlo. The algo used here is described in the textbook:

S. Russell and P. Norvig, "Artificial Intelligence, a Modern Approach,
3rd. Ed." (Prentice Hall, 2010) Section 14.5 "Approximate Inference in
Bayesian Networks", page 537

Attributes
----------
bnet_ord_nodes : list[BayesNode]

bnet : BayesNet
do_print : bool
is_quantum : bool

Constructor & Destructor Documentation

def inference.MCMC_Engine.MCMC_Engine.__init__	(		self,
			bnet,
			do_print = False,
			is_quantum = False
	)

Constructor

Parameters
----------
bnet : BayesNet
do_print : bool
is_quantum : bool

Returns
-------

Member Function Documentation

def inference.MCMC_Engine.MCMC_Engine.get_unipot_list	(		self,
			node_list,
			num_cycles,
			warmup
	)

For each node in node_list, this method returns a uni-potential that
gives the probabilities for the states of that node. Obviously,
such a PD has the active states of the node as support.

Parameters
----------
node_list : list[BayesNode]
num_cycles : int
    How many times we should cycle through all the nodes of the graph
warmup : int
    How many cycles should we wait before we start collecting data
    for the unipot list.

Returns
-------
list[DiscreteUniPot]

def inference.MCMC_Engine.MCMC_Engine.sample_node_given_markov_blanket	(		self,
			focus_node,
			annotated_story
	)

For fixed values of the states of the Markov Blanket of the focus
node, sample a possible state of the focus node. The states of the
Markov Blanket nodes are gleaned from 'annotated_story'.

annotated_story is a dictionary mapping bnet_ord_nodes to their
current state.

A story (aka history or Feynman path or bnet instantiation) is a
list of the current states of the nodes in the list bnet_ord_nodes.

Parameters
----------
focus_node : BayesNode
annotated_story : dict[BayesNode, int]

Returns
-------
(int, DiscreteUniPot)

---

The documentation for this class was generated from the following file:

• inference/MCMC_Engine.py