def learn (self, data, header=None, gamma=0.001, verbose=False): " Estimate parameters from data " start = time.time() # mark starting time if header is None: header = [self.class_variable] + self.attributes learner = ParameterLearner(self.bn, data, header) # header contains the order of appearance of variables in file (default is attribute variables list passed in __init__) learner.learn(gamma) # gamma is the prior probabilities parameter end = time.time() # mark ending time if verbose: print 'Elapsed time: %dm%.3fs' % ((end-start)/60.0,(end-start)%60.0)
def run():
" Run this example "
begining = time.time()
name = "dogproblem"
header = \
""" The Dog Problem Network
F B
|\ /
| \ /
v v
L D-->H
Extracted from E. Charniak.
Bayesian networks without tears.
AI Magazine, 1991. \n """
F = RandomVariable('family_out',['yes','no'])
B = RandomVariable('bowel_problem',['yes','no'])
L = RandomVariable('lights',['on', 'off'])
D = RandomVariable('dog_out',['yes', 'no'])
H = RandomVariable('hear_bark',['yes', 'no'])
### Graph Nodes
V = [F,B,L,D,H]
### Graph Arcs
E = [(F,L), (F,D),
(B,D), (D,H)]
### Bayesian network
g = DBN(V,E,name,header)
markone = time.time() # mark time
dat = load_csv(data_filename)
names = [F,B,L,D,H] # variable ordering in datafile
marktwo = time.time() # mark time
learner = ParameterLearner(g,dat,names)
learner.learn(0.001)
markthree = time.time() # mark time
#for v in g.V:
# print v.cpt
print g
end = time.time()
print 'Elapsed time on network construction: %dm%.3fs' % ((markone-begining)/60.0,(markone-begining)%60.0)
print 'Elapsed time on loading data: %dm%.3fs' % ((marktwo-markone)/60.0,(marktwo-markone)%60.0)
print 'Elapsed time on learning: %dm%.3fs' % ((markthree-marktwo)/60.0,(markthree-marktwo)%60.0)
print 'Total elapsed time: %dm%.3fs' % ((end-begining)/60.0,(end-begining)%60.0)