from Factor import *
from PGMcommon import *
from CliqueTree import *
from CliqueTreeOperations import *
from FactorOperations import *
import scipy.io as sio
import numpy as np
import pprint
import pdb
matfile='/Users/amit/BC_Classes/PGM/Prog4/PA4Sample.mat'
mat_contents=sio.loadmat(matfile)
mat_struct=mat_contents['SumProdCalibrate']
val=mat_struct[0,0]
input_edges = val['INPUT']['edges'][0][0]
input_cliqueList= val['INPUT']['cliqueList'][0][0][0]
clique_list_factorObj=[]
for tpl in input_cliqueList:
(var, card, values)=tpl
f= Factor( var[0].tolist(), card[0].tolist(), values[0].tolist(), 'factor' )
clique_list_factorObj.append(f)
P=CliqueTree( clique_list_factorObj , input_edges, clique_list_factorObj, [])
P=CliqueTreeCalibrate(P)
for f in P.getNodeList():
print f
print "=="
from Factor import *
from PGMcommon import *
from CliqueTree import *
from CliqueTreeOperations import *
from FactorOperations import *
import scipy.io as sio
import numpy as np
import pprint
import pdb
matfile='/Users/amit/BC_Classes/PGM/Prog4/PA4Sample.mat'
mat_contents=sio.loadmat(matfile)
mat_struct=mat_contents['MaxSumCalibrate']
val=mat_struct[0,0]
input_edges = val['INPUT']['edges'][0][0]
#print input_edges
input_cliqueList= val['INPUT']['cliqueList'][0][0][0]
clique_list_factorObj=[]
for tpl in input_cliqueList:
(var, card, values)=tpl
f= Factor( var[0].tolist(), card[0].tolist(), values[0].tolist(), 'factor' )
#print f
clique_list_factorObj.append(f)
P=CliqueTree( clique_list_factorObj , input_edges, clique_list_factorObj, [])
P=CliqueTreeCalibrate(P,1)
for f in P.getNodeList():
print f
print "=="
np.shape(mat_struct)
val = mat_struct[0, 0]
input_edges = val['INPUT1']['edges'][0][0]
input_cliqueList = val['INPUT1']['cliqueList'][0][0][0]
clique_list_factorObj = []
for tpl in input_cliqueList:
(var, card, values) = tpl
f = Factor(var[0].tolist(), card[0].tolist(), values[0].tolist(), 'factor')
clique_list_factorObj.append(f)
messages = val['INPUT2']
messageFactors = []
(nrow, ncol) = np.shape(messages)
for i in range(nrow):
for j in range(ncol):
(var, card, values) = messages[i][j]
f = Factor(var[0].tolist(), card[0].tolist(), values[0].tolist(),
'factor')
messageFactors.append(f)
MESSAGES = np.reshape(np.array(messageFactors), (nrow, ncol))
""" dump it to Python pickle file """
with open('GetNextC.INPUT1.pickle', 'wb') as f:
pickle.dump(clique_list_factorObj, f)
with open('GetNextC.INPUT2.pickle', 'wb') as f:
pickle.dump(MESSAGES, f)
P = CliqueTree(clique_list_factorObj, input_edges, clique_list_factorObj, [])
(a, b) = getNextClique(P, MESSAGES)
print 'a: ', a, ' b:', b
def createCliqueTree(factorList, E=[]):
""" return a Clique Tree object given a list of factors
it peforms VE and returns the clique tree the VE
ordering defines. See Chapter 9 of Friedman and Koller
Probabilistic Graphical Models"""
V = getUniqueVar(factorList)
totalVars = len(V)
cardinality = np.zeros(len(V)).tolist()
for i in range(len(V)):
for j in range(len(factorList)):
try:
indx = factorList[j].getVar().tolist().index(V[i])
cardinality[i] = factorList[j].getCard().tolist()[indx]
break
except:
continue
edges = np.zeros((totalVars, totalVars))
""" Set up adjacency matrix: for each factor, get the list of variables in its scope and create an edge between each variable in the factor """
for f in factorList:
variableList = f.getVar()
for j in range(len(variableList)):
for k in range(len(variableList)):
edges[variableList[j] - 1, variableList[k] - 1] = 1
(nrows, nedges) = np.shape(edges)
C = CliqueTree()
C.setCard(cardinality)
C.setEdges(np.zeros((totalVars, totalVars)))
C.setFactorList(factorList)
C.setEvidence(E)
C.setNodeList([])
#print 'length of factorList: ', len(factorList)
#print C.toString()
cliquesConsidered = 0
#pdb.set_trace()
while cliquesConsidered < len(V):
bestClique = 0
bestScore = sys.maxint
for i in range(nrows):
score = np.sum(edges[i, :])
if score > 0 and score < bestScore:
bestScore = score
bestClique = i + 1
cliquesConsidered += 1
(edges, factorList) = C.eliminateVar(bestClique, edges, factorList)
return C
def createCliqueTree( factorList,E=[]):
""" return a Clique Tree object given a list of factors
it peforms VE and returns the clique tree the VE
ordering defines. See Chapter 9 of Friedman and Koller
Probabilistic Graphical Models"""
V=getUniqueVar(factorList)
totalVars=len(V)
cardinality=np.zeros(len(V)).tolist()
for i in range(len(V)):
for j in range(len(factorList)):
try:
indx= factorList[j].getVar().tolist().index( V[i] )
cardinality[i]=factorList[j].getCard().tolist()[indx]
break
except:
continue
edges=np.zeros( (totalVars, totalVars))
""" Set up adjacency matrix: for each factor, get the list of variables in its scope and create an edge between each variable in the factor """
for f in factorList:
variableList=f.getVar()
for j in range(len(variableList) ):
for k in range (len(variableList) ):
edges[ variableList[j]-1, variableList[k]-1 ]=1
(nrows,nedges)=np.shape(edges)
C=CliqueTree()
C.setCard( cardinality )
C.setEdges(np.zeros( (totalVars, totalVars)))
C.setFactorList(factorList)
C.setEvidence(E)
C.setNodeList([])
#print 'length of factorList: ', len(factorList)
#print C.toString()
cliquesConsidered = 0
#pdb.set_trace()
while cliquesConsidered < len(V):
bestClique = 0
bestScore = sys.maxint
for i in range(nrows):
score=np.sum( edges[i,:] )
if score > 0 and score < bestScore:
bestScore = score
bestClique = i+1
cliquesConsidered+=1
(edges, factorList)=C.eliminateVar(bestClique, edges, factorList)
return C
