def test_sample_binary_random_variable():
cpt = {'T': {'T': 1.0, 'F': 0.0}, 'F': {'T': 0.0, 'F': 1.0}, 'prior': {'T': 1.0, 'F': 0.0}}
for i in range(100):
s = problem_1.sample_binary_random_variable('prior', cpt)
assert s == 'T'
s = problem_1.sample_binary_random_variable('T', cpt)
assert s == 'T'
s = problem_1.sample_binary_random_variable('F', cpt)
assert s == 'F'
def particle_filtering_umbrella_network(numSamples, umbrella_evidence, rain_cpt, umbrella_cpt):
# S is a list of N particles
# each particle is 'T' or 'F'
# initialize the particles with the prior probability of rain P(rain=T)=0.5
S = [problem_1.sample_binary_random_variable('prior', rain_cpt) for n in range(numSamples)]
W = [1.0] * numSamples
for i, e in enumerate(umbrella_evidence):
for n in range(numSamples):
S[n] = problem_1.sample_binary_random_variable(S[n], rain_cpt)
W[n] *= umbrella_cpt[S[n]][e]
S, W = weighted_sample_with_replacement(S, W)
return S
def weighted_sample_umbrella_network(umbrella_evidence, rain_cpt, umbrella_cpt):
# return sample x and weight w
x = []
w = 1.0
# sample rain for t=0
rain_0 = problem_1.sample_binary_random_variable('prior', rain_cpt)
w *= weight_for_evidence(rain_0, umbrella_cpt, umbrella_evidence[0])
x.append(rain_0)
# sample rain for t>0
for t in range(1, len(umbrella_evidence)):
# sample rain_t
rain_t_minus_1 = x[t - 1]
rain_t = problem_1.sample_binary_random_variable(rain_t_minus_1, rain_cpt)
x.append(rain_t)
# update the weight
w *= weight_for_evidence(rain_t, umbrella_cpt, umbrella_evidence[t])
return x, w
