def testReportObserveInferActionSmoke():
vals = get_ripl().execute_program("""\
[assume x (normal 0 1)]
foo : [observe (normal x 1) 0.2]
[infer (mh default one 1)]
(report 'foo)
""")
eq_(0.2, u.strip_types(vals[3]['value']))
def testReportObserveListActionSmoke():
vals = get_ripl().execute_program("""\
[assume m (array 0. 0.)]
[assume s (matrix (array (array 1. 0.) (array 0. 1.)))]
foo : [observe (multivariate_normal m s) (array 0.1 -0.1)]
(report 'foo)
""")
eq_([0.1, -0.1], u.strip_types(vals[3]['value']))
def testForceSugar():
r = get_ripl()
r.set_mode("venture_script")
vals = r.execute_program("""\
assume x = normal(0,1);
force x = 5;
report(quote(x));
""")
eq_(5, u.strip_types([v['value'] for v in vals])[2])
def hmm_map(inferrer, states):
sequence = map(int, u.strip_types(states[0]))
time = hmm_time()
hmm_pause()
import sys
print >> sys.stderr, 'MAP sequence: %s' % (sequence, )
print >> sys.stderr, 'Score: %s' % inferrer.engine.ripl.infer(
'global_log_joint')[0]
print '%s,%s' % (time, metric1(sequence))
hmm_resume()
def testEnumerativeSmoke():
r = get_ripl()
r.assume("x", "(categorical (simplex 0.1 0.2 0.3 0.4) (list 1 2 3 4))")
r.infer("(enumerative_diversify default all)")
assert np.allclose([1, 2, 3, 4],
strip_types(
r.sivm.core_sivm.engine.sample_all(v.sym("x"))))
assert np.allclose([0.1, 0.2, 0.3, 0.4],
logWeightsToNormalizedDirect(
r.sivm.core_sivm.engine.model.log_weights))
def hmm_state(inferrer, states):
global hmm_marginals
assert hmm_marginals is not None
import sys
marginals = marginalize(inferrer, u.strip_types(states))
tvd = total_variation_distance(marginals,
filtered_marginals[len(hmm_marginals)])
print >> sys.stderr, ('Filtered marginals at %2d (tvd %s): %s' % (
len(hmm_marginals),
tvd,
marginals,
))
hmm_marginals.append(marginals)
def testEnumerativeStacking():
r = get_ripl()
r.assume("x", "(categorical (simplex 0.1 0.2 0.3 0.4) (list 1 2 3 4))")
r.infer("(enumerative_diversify default all)")
r.infer("(enumerative_diversify default all)")
assert np.allclose([1, 2, 3, 4] * 4,
strip_types(
r.sivm.core_sivm.engine.sample_all(v.sym("x"))))
assert np.allclose(
[
0.01, 0.02, 0.03, 0.04, 0.02, 0.04, 0.06, 0.08, 0.03, 0.06, 0.09,
0.12, 0.04, 0.08, 0.12, 0.16
], #TODO Are these actually the weights I want here?
logWeightsToNormalizedDirect(
r.sivm.core_sivm.engine.model.log_weights))
def hmm_smoothed(inferrer, states, likelihood_weight=True):
states = u.strip_types(states)
n = len(states[0])
marginals = [
marginalize(inferrer, [s[i] for s in states], likelihood_weight)
for i in range(n)
]
time = hmm_time()
hmm_pause()
for i in range(n):
import sys
tvd = total_variation_distance(marginals[i], smoothed_marginals[i])
print >> sys.stderr, ('Smoothed marginals at %2d (tvd %s): %s' %
(i, tvd, marginals[i]))
m, v = metric2(marginals)
print '%s,%s,%s' % (time, m, v)
hmm_resume()
def printValue(directive):
'''Gets the actual value returned by an assume, predict, report, or sample directive.'''
if directive is None or 'value' not in directive:
pass
else:
print strip_types(directive['value'])
def testSampleSugar():
r = get_ripl()
r.set_mode("venture_script")
vals = r.execute_program("sample 2 + 2;")
eq_([4], u.strip_types([v['value'] for v in vals]))
def testReportObserveActionSmoke():
vals = get_ripl().execute_program("""\
foo : [observe (normal 0 1) 0.2]
(report 'foo)
""")
eq_(0.2, u.strip_types(vals[1]['value']))
def testReportActionSmoke():
vals = get_ripl().execute_program("""\
foo : [assume x (+ 1 2)]
(report 'foo)
""")
eq_([3.0, 3.0], u.strip_types([v['value'] for v in vals]))
