def test_crosscat_constraints():
class FakeEngine(crosscat.LocalEngine.LocalEngine):
def predictive_probability_multistate(self, M_c, X_L_list, X_D_list, Y,
Q):
self._last_Y = Y
sup = super(FakeEngine, self)
return sup.simple_predictive_probability_multistate(
M_c=M_c, X_L_list=X_L_list, X_D_list=X_D_list, Y=Y, Q=Q)
def simple_predictive_sample(self, seed, M_c, X_L, X_D, Y, Q, n):
self._last_Y = Y
return super(FakeEngine, self).simple_predictive_sample(seed=seed,
M_c=M_c,
X_L=X_L,
X_D=X_D,
Y=Y,
Q=Q,
n=n)
def impute_and_confidence(self, seed, M_c, X_L, X_D, Y, Q, n):
self._last_Y = Y
return super(FakeEngine, self).impute_and_confidence(seed=seed,
M_c=M_c,
X_L=X_L,
X_D=X_D,
Y=Y,
Q=Q,
n=n)
engine = FakeEngine(seed=0)
mm = CrosscatMetamodel(engine)
with bayesdb(metamodel=mm) as bdb:
t1_schema(bdb)
t1_data(bdb)
bdb.execute('''
CREATE GENERATOR t1_cc FOR t1 USING crosscat(
label CATEGORICAL,
age NUMERICAL,
weight NUMERICAL
)
''')
gid = core.bayesdb_get_generator(bdb, 't1_cc')
assert core.bayesdb_generator_column_number(bdb, gid, 'label') == 1
assert core.bayesdb_generator_column_number(bdb, gid, 'age') == 2
assert core.bayesdb_generator_column_number(bdb, gid, 'weight') == 3
from bayeslite.metamodels.crosscat import crosscat_cc_colno
assert crosscat_cc_colno(bdb, gid, 1) == 0
assert crosscat_cc_colno(bdb, gid, 2) == 1
assert crosscat_cc_colno(bdb, gid, 3) == 2
bdb.execute('INITIALIZE 1 MODEL FOR t1_cc')
bdb.execute('ANALYZE t1_cc FOR 1 ITERATION WAIT')
bdb.execute('ESTIMATE PROBABILITY OF age = 8 GIVEN (weight = 16)'
' BY t1_cc').next()
assert engine._last_Y == [(28, 2, 16)]
bdb.execute("SELECT age FROM t1 WHERE label = 'baz'").next()
bdb.execute("INFER age FROM t1_cc WHERE label = 'baz'").next()
assert engine._last_Y == [(3, 0, 1), (3, 2, 32)]
bdb.execute('SIMULATE weight FROM t1_cc GIVEN age = 8 LIMIT 1').next()
assert engine._last_Y == [(28, 1, 8)]
def test_crosscat_constraints():
class FakeEngine(crosscat.LocalEngine.LocalEngine):
def predictive_probability_multistate(self, M_c, X_L_list,
X_D_list, Y, Q):
self._last_Y = Y
sup = super(FakeEngine, self)
return sup.simple_predictive_probability_multistate(M_c=M_c,
X_L_list=X_L_list, X_D_list=X_D_list, Y=Y, Q=Q)
def simple_predictive_sample(self, seed, M_c, X_L, X_D, Y, Q, n):
self._last_Y = Y
return super(FakeEngine, self).simple_predictive_sample(seed=seed,
M_c=M_c, X_L=X_L, X_D=X_D, Y=Y, Q=Q, n=n)
def impute_and_confidence(self, seed, M_c, X_L, X_D, Y, Q, n):
self._last_Y = Y
return super(FakeEngine, self).impute_and_confidence(seed=seed,
M_c=M_c, X_L=X_L, X_D=X_D, Y=Y, Q=Q, n=n)
engine = FakeEngine(seed=0)
mm = CrosscatMetamodel(engine)
with bayesdb(metamodel=mm) as bdb:
t1_schema(bdb)
t1_data(bdb)
bdb.execute('''
CREATE POPULATION p1 FOR t1 (
id IGNORE;
label CATEGORICAL;
age NUMERICAL;
weight NUMERICAL
)
''')
bdb.execute('''
CREATE GENERATOR p1_cc FOR p1 USING crosscat(
label CATEGORICAL,
age NUMERICAL,
weight NUMERICAL
)
''')
pid = core.bayesdb_get_population(bdb, 'p1')
assert core.bayesdb_variable_number(bdb, pid, None, 'label') == 1
assert core.bayesdb_variable_number(bdb, pid, None, 'age') == 2
assert core.bayesdb_variable_number(bdb, pid, None, 'weight') == 3
gid = core.bayesdb_get_generator(bdb, pid, 'p1_cc')
from bayeslite.metamodels.crosscat import crosscat_cc_colno
assert crosscat_cc_colno(bdb, gid, 1) == 0
assert crosscat_cc_colno(bdb, gid, 2) == 1
assert crosscat_cc_colno(bdb, gid, 3) == 2
bdb.execute('INITIALIZE 1 MODEL FOR p1_cc')
bdb.execute('ANALYZE p1_cc FOR 1 ITERATION WAIT')
bdb.execute('ESTIMATE PROBABILITY DENSITY OF age = 8'
' GIVEN (weight = 16)'
' BY p1').next()
assert engine._last_Y == [(28, 2, 16)]
bdb.execute("SELECT age FROM t1 WHERE label = 'baz'").next()
bdb.execute("INFER age FROM p1 WHERE label = 'baz'").next()
assert engine._last_Y == [(3, 0, 1), (3, 2, 32)]
bdb.execute('SIMULATE weight FROM p1 GIVEN age = 8 LIMIT 1').next()
assert engine._last_Y == [(28, 1, 8)]
# Simulate with an unknown nominal value should throw an error.
with pytest.raises(bayeslite.BQLError):
bdb.execute('SIMULATE weight FROM p1 GIVEN label = \'q\' LIMIT 1;')
def test_crosscat_constraints():
class FakeEngine(crosscat.LocalEngine.LocalEngine):
def predictive_probability_multistate(self, M_c, X_L_list, X_D_list, Y, Q):
self._last_Y = Y
sup = super(FakeEngine, self)
return sup.simple_predictive_probability_multistate(M_c=M_c, X_L_list=X_L_list, X_D_list=X_D_list, Y=Y, Q=Q)
def simple_predictive_sample(self, M_c, X_L, X_D, Y, Q, n):
self._last_Y = Y
return super(FakeEngine, self).simple_predictive_sample(M_c=M_c, X_L=X_L, X_D=X_D, Y=Y, Q=Q, n=n)
def impute_and_confidence(self, M_c, X_L, X_D, Y, Q, n):
self._last_Y = Y
return super(FakeEngine, self).impute_and_confidence(M_c=M_c, X_L=X_L, X_D=X_D, Y=Y, Q=Q, n=n)
engine = FakeEngine(seed=0)
mm = CrosscatMetamodel(engine)
with bayesdb(metamodel=mm) as bdb:
t1_schema(bdb)
t1_data(bdb)
bdb.execute(
"""
CREATE GENERATOR t1_cc FOR t1 USING crosscat(
label CATEGORICAL,
age NUMERICAL,
weight NUMERICAL
)
"""
)
gid = core.bayesdb_get_generator(bdb, "t1_cc")
assert core.bayesdb_generator_column_number(bdb, gid, "label") == 1
assert core.bayesdb_generator_column_number(bdb, gid, "age") == 2
assert core.bayesdb_generator_column_number(bdb, gid, "weight") == 3
from bayeslite.metamodels.crosscat import crosscat_cc_colno
assert crosscat_cc_colno(bdb, gid, 1) == 0
assert crosscat_cc_colno(bdb, gid, 2) == 1
assert crosscat_cc_colno(bdb, gid, 3) == 2
bdb.execute("INITIALIZE 1 MODEL FOR t1_cc")
bdb.execute("ANALYZE t1_cc FOR 1 ITERATION WAIT")
bdb.execute("ESTIMATE PROBABILITY OF age = 8 GIVEN (weight = 16)" " BY t1_cc").next()
assert engine._last_Y == [(28, 2, 16)]
bdb.execute("SELECT age FROM t1 WHERE label = 'baz'").next()
bdb.execute("INFER age FROM t1_cc WHERE label = 'baz'").next()
assert engine._last_Y == [(3, 0, 1), (3, 2, 32)]
bdb.execute("SIMULATE weight FROM t1_cc GIVEN age = 8 LIMIT 1").next()
assert engine._last_Y == [(28, 1, 8)]