def test_warm_start(self):
# create metamodel with warm_restart = True
meta = MetaModel()
meta.add_param('x1', 0.)
meta.add_param('x2', 0.)
meta.add_output('y1', 0.)
meta.add_output('y2', 0.)
meta.default_surrogate = ResponseSurface()
meta.warm_restart = True
# add to problem
prob = Problem(Group())
prob.root.add('meta', meta)
prob.setup(check=False)
# provide initial training data
prob['meta.train:x1'] = [1.0, 3.0]
prob['meta.train:x2'] = [1.0, 4.0]
prob['meta.train:y1'] = [3.0, 1.0]
prob['meta.train:y2'] = [1.0, 7.0]
# run against a data point and check result
prob['meta.x1'] = 2.0
prob['meta.x2'] = 3.0
prob.run()
assert_rel_error(self, prob['meta.y1'], 1.9085, .001)
assert_rel_error(self, prob['meta.y2'], 3.9203, .001)
# Add 3rd training point, moves the estimate for that point
# back to where it should be.
prob['meta.train:x1'] = [2.0]
prob['meta.train:x2'] = [3.0]
prob['meta.train:y1'] = [2.0]
prob['meta.train:y2'] = [4.0]
meta.train = True # currently need to tell meta to re-train
prob.run()
assert_rel_error(self, prob['meta.y1'], 2.0, .00001)
assert_rel_error(self, prob['meta.y2'], 4.0, .00001)
def test_sin_metamodel(self):
class Sin(Component):
""" Simple sine calculation. """
def __init__(self):
self.add_param('x', 0., units="rad")
self.add_output('f_x', 0.)
def solve_nonlinear(self, params, unknowns, resids):
unknowns['f_x'] = .5*sin(params['x'])
# create a MetaModel for Sin and add it to a Problem
sin_mm = MetaModel()
sin_mm.add_param('x', 0.)
sin_mm.add_output('f_x', 0.)
prob = Problem(Group())
prob.root.add('sin_mm', sin_mm)
# check that missing surrogate is detected in check_setup
stream = cStringIO()
prob.setup(out_stream=stream)
msg = ("No default surrogate model is defined and the "
"following outputs do not have a surrogate model:\n"
"['f_x']\n"
"Either specify a default_surrogate, or specify a "
"surrogate model for all outputs.")
self.assertTrue(msg in stream.getvalue())
# check that output with no specified surrogate gets the default
sin_mm.default_surrogate = FloatKrigingSurrogate()
prob.setup(check=False)
surrogate = prob.root.unknowns.metadata('sin_mm.f_x').get('surrogate')
self.assertTrue(isinstance(surrogate, FloatKrigingSurrogate),
'sin_mm.f_x should get the default surrogate')
# train the surrogate and check predicted value
prob['sin_mm.train:x'] = np.linspace(0,10,200)
prob['sin_mm.train:f_x'] = .5*np.sin(prob['sin_mm.train:x'])
prob['sin_mm.x'] = 2.22
prob.run()
self.assertAlmostEqual(prob['sin_mm.f_x'],
.5*np.sin(prob['sin_mm.x']),
places=5)
def test_basics(self):
# create a metamodel component
mm = MetaModel()
mm.add_param('x1', 0.)
mm.add_param('x2', 0.)
mm.add_output('y1', 0.)
mm.add_output('y2', 0., surrogate=FloatKrigingSurrogate())
mm.default_surrogate = ResponseSurface()
# add metamodel to a problem
prob = Problem(root=Group())
prob.root.add('mm', mm)
prob.setup(check=False)
# check that surrogates were properly assigned
surrogate = prob.root.unknowns.metadata('mm.y1').get('surrogate')
self.assertTrue(isinstance(surrogate, ResponseSurface))
surrogate = prob.root.unknowns.metadata('mm.y2').get('surrogate')
self.assertTrue(isinstance(surrogate, FloatKrigingSurrogate))
# populate training data
prob['mm.train:x1'] = [1.0, 2.0, 3.0]
prob['mm.train:x2'] = [1.0, 3.0, 4.0]
prob['mm.train:y1'] = [3.0, 2.0, 1.0]
prob['mm.train:y2'] = [1.0, 4.0, 7.0]
# run problem for provided data point and check prediction
prob['mm.x1'] = 2.0
prob['mm.x2'] = 3.0
self.assertTrue(mm.train) # training will occur before 1st run
prob.run()
assert_rel_error(self, prob['mm.y1'], 2.0, .00001)
assert_rel_error(self, prob['mm.y2'], 4.0, .00001)
# run problem for interpolated data point and check prediction
prob['mm.x1'] = 2.5
prob['mm.x2'] = 3.5
self.assertFalse(mm.train) # training will not occur before 2nd run
prob.run()
assert_rel_error(self, prob['mm.y1'], 1.5934, .001)
# change default surrogate, re-setup and check that metamodel re-trains
mm.default_surrogate = FloatKrigingSurrogate()
prob.setup(check=False)
surrogate = prob.root.unknowns.metadata('mm.y1').get('surrogate')
self.assertTrue(isinstance(surrogate, FloatKrigingSurrogate))
self.assertTrue(mm.train) # training will occur after re-setup
mm.warm_restart = True # use existing training data
prob['mm.x1'] = 2.5
prob['mm.x2'] = 3.5
prob.run()
assert_rel_error(self, prob['mm.y1'], 1.4609, .001)
