Commit f75a49a8 authored by Gordian Edenhofer's avatar Gordian Edenhofer
Browse files

Swap order of arguments of ScalingOperator

Make the ScalingOperator consistent with `from_global_data`, `full` and
possibly other operators which take the domain as first argument. This
breaks the currently enforced consistency with the `DiagonalOperator`
which takes a field as first argument and the domain only as an optional
keyword argument.

Fixes #279.
parent 4ca56a07
Pipeline #65020 passed with stages
in 8 minutes and 54 seconds
...@@ -240,7 +240,7 @@ ...@@ -240,7 +240,7 @@
"sh = Sh.draw_sample()\n", "sh = Sh.draw_sample()\n",
"noiseless_data=R(sh)\n", "noiseless_data=R(sh)\n",
"noise_amplitude = np.sqrt(0.2)\n", "noise_amplitude = np.sqrt(0.2)\n",
"N = ift.ScalingOperator(noise_amplitude**2, s_space)\n", "N = ift.ScalingOperator(s_space, noise_amplitude**2)\n",
"\n", "\n",
"n = ift.Field.from_random(domain=s_space, random_type='normal',\n", "n = ift.Field.from_random(domain=s_space, random_type='normal',\n",
" std=noise_amplitude, mean=0)\n", " std=noise_amplitude, mean=0)\n",
...@@ -391,7 +391,7 @@ ...@@ -391,7 +391,7 @@
"source": [ "source": [
"# Operators\n", "# Operators\n",
"Sh = ift.create_power_operator(h_space, power_spectrum=pow_spec)\n", "Sh = ift.create_power_operator(h_space, power_spectrum=pow_spec)\n",
"N = ift.ScalingOperator(noise_amplitude**2,s_space)\n", "N = ift.ScalingOperator(s_space, noise_amplitude**2)\n",
"# R is defined below\n", "# R is defined below\n",
"\n", "\n",
"# Fields\n", "# Fields\n",
...@@ -569,7 +569,7 @@ ...@@ -569,7 +569,7 @@
"\n", "\n",
"# Operators\n", "# Operators\n",
"Sh = ift.create_power_operator(h_space, power_spectrum=pow_spec)\n", "Sh = ift.create_power_operator(h_space, power_spectrum=pow_spec)\n",
"N = ift.ScalingOperator(sigma2,s_space)\n", "N = ift.ScalingOperator(s_space, sigma2)\n",
"\n", "\n",
"# Fields and data\n", "# Fields and data\n",
"sh = Sh.draw_sample()\n", "sh = Sh.draw_sample()\n",
......
...@@ -113,7 +113,7 @@ if __name__ == '__main__': ...@@ -113,7 +113,7 @@ if __name__ == '__main__':
# Set the noise covariance N # Set the noise covariance N
noise = 5. noise = 5.
N = ift.ScalingOperator(noise, data_space) N = ift.ScalingOperator(data_space, noise)
# Create mock data # Create mock data
MOCK_SIGNAL = S.draw_sample() MOCK_SIGNAL = S.draw_sample()
......
...@@ -74,7 +74,7 @@ if __name__ == '__main__': ...@@ -74,7 +74,7 @@ if __name__ == '__main__':
# Specify noise # Specify noise
data_space = R.target data_space = R.target
noise = .001 noise = .001
N = ift.ScalingOperator(noise, data_space) N = ift.ScalingOperator(data_space, noise)
# Generate mock signal and data # Generate mock signal and data
mock_position = ift.from_random('normal', signal_response.domain) mock_position = ift.from_random('normal', signal_response.domain)
......
...@@ -93,7 +93,7 @@ if __name__ == '__main__': ...@@ -93,7 +93,7 @@ if __name__ == '__main__':
# Specify noise # Specify noise
data_space = R.target data_space = R.target
noise = .001 noise = .001
N = ift.ScalingOperator(noise, data_space) N = ift.ScalingOperator(data_space, noise)
# Generate mock signal and data # Generate mock signal and data
mock_position = ift.from_random('normal', signal_response.domain) mock_position = ift.from_random('normal', signal_response.domain)
......
...@@ -72,7 +72,7 @@ def make_adjust_variances_hamiltonian(a, ...@@ -72,7 +72,7 @@ def make_adjust_variances_hamiltonian(a,
x = (a.conjugate()*a).real x = (a.conjugate()*a).real
if scaling is not None: if scaling is not None:
x = ScalingOperator(scaling, x.target)(x) x = ScalingOperator(x.target, scaling)(x)
return StandardHamiltonian(InverseGammaLikelihood(d_eval/2.)(x), return StandardHamiltonian(InverseGammaLikelihood(d_eval/2.)(x),
ic_samp=ic_samp) ic_samp=ic_samp)
......
...@@ -74,7 +74,7 @@ def _make_dynamic_operator(target, ...@@ -74,7 +74,7 @@ def _make_dynamic_operator(target,
ops['FFT'] = FFT ops['FFT'] = FFT
ops['Real'] = Real ops['Real'] = Real
if harmonic_padding is None: if harmonic_padding is None:
CentralPadd = ScalingOperator(1., FFT.target) CentralPadd = ScalingOperator(FFT.target, 1.)
else: else:
if isinstance(harmonic_padding, int): if isinstance(harmonic_padding, int):
harmonic_padding = list((harmonic_padding,)*len(FFT.target.shape)) harmonic_padding = list((harmonic_padding,)*len(FFT.target.shape))
...@@ -123,7 +123,7 @@ def _make_dynamic_operator(target, ...@@ -123,7 +123,7 @@ def _make_dynamic_operator(target,
c = FieldAdapter(UnstructuredDomain(len(sigc)), keys[1]) c = FieldAdapter(UnstructuredDomain(len(sigc)), keys[1])
c = makeOp(Field.from_global_data(c.target, np.array(sigc)))(c) c = makeOp(Field.from_global_data(c.target, np.array(sigc)))(c)
lightspeed = ScalingOperator(-0.5, c.target)(c).exp() lightspeed = ScalingOperator(c.target, -0.5)(c).exp()
scaling = np.array(m.target[0].distances[1:])/m.target[0].distances[0] scaling = np.array(m.target[0].distances[1:])/m.target[0].distances[0]
scaling = DiagonalOperator(Field.from_global_data(c.target, scaling)) scaling = DiagonalOperator(Field.from_global_data(c.target, scaling))
ops['lightspeed'] = scaling(lightspeed) ops['lightspeed'] = scaling(lightspeed)
......
...@@ -406,7 +406,7 @@ class Linearization(object): ...@@ -406,7 +406,7 @@ class Linearization(object):
the requested Linearization the requested Linearization
""" """
from .operators.scaling_operator import ScalingOperator from .operators.scaling_operator import ScalingOperator
return Linearization(field, ScalingOperator(1., field.domain), return Linearization(field, ScalingOperator(field.domain, 1.),
want_metric=want_metric) want_metric=want_metric)
@staticmethod @staticmethod
...@@ -492,7 +492,7 @@ class Linearization(object): ...@@ -492,7 +492,7 @@ class Linearization(object):
if len(constants) == 0: if len(constants) == 0:
return Linearization.make_var(field, want_metric) return Linearization.make_var(field, want_metric)
else: else:
ops = {key: ScalingOperator(0. if key in constants else 1., dom) ops = {key: ScalingOperator(dom, 0. if key in constants else 1.)
for key, dom in field.domain.items()} for key, dom in field.domain.items()}
bdop = BlockDiagonalOperator(field.domain, ops) bdop = BlockDiagonalOperator(field.domain, ops)
return Linearization(field, bdop, want_metric=want_metric) return Linearization(field, bdop, want_metric=want_metric)
...@@ -72,7 +72,7 @@ class ChainOperator(LinearOperator): ...@@ -72,7 +72,7 @@ class ChainOperator(LinearOperator):
break break
if fct != 1 or len(opsnew) == 0: if fct != 1 or len(opsnew) == 0:
# have to add the scaling operator at the end # have to add the scaling operator at the end
opsnew.append(ScalingOperator(fct, lastdom)) opsnew.append(ScalingOperator(lastdom, fct))
ops = opsnew ops = opsnew
# combine DiagonalOperators where possible # combine DiagonalOperators where possible
opsnew = [] opsnew = []
......
...@@ -277,7 +277,7 @@ class StudentTEnergy(EnergyOperator): ...@@ -277,7 +277,7 @@ class StudentTEnergy(EnergyOperator):
return Field.scalar(v) return Field.scalar(v)
if not x.want_metric: if not x.want_metric:
return v return v
met = ScalingOperator((self._theta+1)/(self._theta+3), self.domain) met = ScalingOperator(self.domain, (self._theta+1) / (self._theta+3))
met = SandwichOperator.make(x.jac, met) met = SandwichOperator.make(x.jac, met)
return v.add_metric(met) return v.add_metric(met)
......
...@@ -417,7 +417,7 @@ def HarmonicSmoothingOperator(domain, sigma, space=None): ...@@ -417,7 +417,7 @@ def HarmonicSmoothingOperator(domain, sigma, space=None):
if sigma < 0.: if sigma < 0.:
raise ValueError("sigma must be non-negative") raise ValueError("sigma must be non-negative")
if sigma == 0.: if sigma == 0.:
return ScalingOperator(1., domain) return ScalingOperator(domain, 1.)
domain = DomainTuple.make(domain) domain = DomainTuple.make(domain)
space = utilities.infer_space(domain, space) space = utilities.infer_space(domain, space)
......
...@@ -60,7 +60,7 @@ class Operator(metaclass=NiftyMeta): ...@@ -60,7 +60,7 @@ class Operator(metaclass=NiftyMeta):
if factor == 1: if factor == 1:
return self return self
from .scaling_operator import ScalingOperator from .scaling_operator import ScalingOperator
return ScalingOperator(factor, self.target)(self) return ScalingOperator(self.target, factor)(self)
def conjugate(self): def conjugate(self):
from .simple_linear_operators import ConjugationOperator from .simple_linear_operators import ConjugationOperator
......
...@@ -57,7 +57,7 @@ class SandwichOperator(EndomorphicOperator): ...@@ -57,7 +57,7 @@ class SandwichOperator(EndomorphicOperator):
if cheese is not None and not isinstance(cheese, LinearOperator): if cheese is not None and not isinstance(cheese, LinearOperator):
raise TypeError("cheese must be a linear operator or None") raise TypeError("cheese must be a linear operator or None")
if cheese is None: if cheese is None:
cheese = ScalingOperator(1., bun.target) cheese = ScalingOperator(bun.target, 1.)
op = bun.adjoint(bun) op = bun.adjoint(bun)
else: else:
op = bun.adjoint(cheese(bun)) op = bun.adjoint(cheese(bun))
......
...@@ -26,10 +26,10 @@ class ScalingOperator(EndomorphicOperator): ...@@ -26,10 +26,10 @@ class ScalingOperator(EndomorphicOperator):
Parameters Parameters
---------- ----------
factor : scalar
The multiplication factor
domain : Domain or tuple of Domain or DomainTuple domain : Domain or tuple of Domain or DomainTuple
The domain on which the Operator's input Field is defined. The domain on which the Operator's input Field is defined.
factor : scalar
The multiplication factor
Notes Notes
----- -----
...@@ -50,13 +50,13 @@ class ScalingOperator(EndomorphicOperator): ...@@ -50,13 +50,13 @@ class ScalingOperator(EndomorphicOperator):
somewhere else. somewhere else.
""" """
def __init__(self, factor, domain): def __init__(self, domain, factor):
from ..sugar import makeDomain from ..sugar import makeDomain
if not np.isscalar(factor): if not np.isscalar(factor):
raise TypeError("Scalar required") raise TypeError("Scalar required")
self._factor = factor
self._domain = makeDomain(domain) self._domain = makeDomain(domain)
self._factor = factor
self._capability = self._all_ops self._capability = self._all_ops
def apply(self, x, mode): def apply(self, x, mode):
...@@ -81,7 +81,7 @@ class ScalingOperator(EndomorphicOperator): ...@@ -81,7 +81,7 @@ class ScalingOperator(EndomorphicOperator):
fct = np.conj(fct) fct = np.conj(fct)
if trafo & self.INVERSE_BIT: if trafo & self.INVERSE_BIT:
fct = 1./fct fct = 1./fct
return ScalingOperator(fct, self._domain) return ScalingOperator(self._domain, fct)
def _get_fct(self, from_inverse): def _get_fct(self, from_inverse):
fct = self._factor fct = self._factor
......
...@@ -99,7 +99,7 @@ class SumOperator(LinearOperator): ...@@ -99,7 +99,7 @@ class SumOperator(LinearOperator):
break break
if sum != 0 or len(opsnew) == 0: if sum != 0 or len(opsnew) == 0:
# have to add the scaling operator at the end # have to add the scaling operator at the end
opsnew.append(ScalingOperator(sum, lastdom)) opsnew.append(ScalingOperator(lastdom, sum))
negnew.append(False) negnew.append(False)
ops = opsnew ops = opsnew
......
...@@ -518,7 +518,7 @@ def calculate_position(operator, output): ...@@ -518,7 +518,7 @@ def calculate_position(operator, output):
if output.domain != operator.target: if output.domain != operator.target:
raise TypeError raise TypeError
cov = 1e-3*output.to_global_data().max()**2 cov = 1e-3*output.to_global_data().max()**2
invcov = ScalingOperator(cov, output.domain).inverse invcov = ScalingOperator(output.domain, cov).inverse
d = output + invcov.draw_sample(from_inverse=True) d = output + invcov.draw_sample(from_inverse=True)
lh = GaussianEnergy(d, invcov)(operator) lh = GaussianEnergy(d, invcov)(operator)
H = StandardHamiltonian( H = StandardHamiltonian(
......
...@@ -37,7 +37,7 @@ PARAMS = product(SEEDS, SPACES) ...@@ -37,7 +37,7 @@ PARAMS = product(SEEDS, SPACES)
@pytest.fixture(params=PARAMS) @pytest.fixture(params=PARAMS)
def field(request): def field(request):
np.random.seed(request.param[0]) np.random.seed(request.param[0])
S = ift.ScalingOperator(1., request.param[1]) S = ift.ScalingOperator(request.param[1], 1.)
s = S.draw_sample() s = S.draw_sample()
return ift.MultiField.from_dict({'s1': s})['s1'] return ift.MultiField.from_dict({'s1': s})['s1']
...@@ -76,7 +76,7 @@ def test_hamiltonian_and_KL(field): ...@@ -76,7 +76,7 @@ def test_hamiltonian_and_KL(field):
lh = ift.GaussianEnergy(domain=space) lh = ift.GaussianEnergy(domain=space)
hamiltonian = ift.StandardHamiltonian(lh) hamiltonian = ift.StandardHamiltonian(lh)
ift.extra.check_jacobian_consistency(hamiltonian, field) ift.extra.check_jacobian_consistency(hamiltonian, field)
S = ift.ScalingOperator(1., space) S = ift.ScalingOperator(space, 1.)
samps = [S.draw_sample() for i in range(3)] samps = [S.draw_sample() for i in range(3)]
kl = ift.AveragedEnergy(hamiltonian, samps) kl = ift.AveragedEnergy(hamiltonian, samps)
ift.extra.check_jacobian_consistency(kl, field) ift.extra.check_jacobian_consistency(kl, field)
......
...@@ -51,9 +51,9 @@ def test_gaussian_energy(space, nonlinearity, noise, seed): ...@@ -51,9 +51,9 @@ def test_gaussian_energy(space, nonlinearity, noise, seed):
pspec = ift.PS_field(pspace, pspec) pspec = ift.PS_field(pspace, pspec)
A = Dist(ift.sqrt(pspec)) A = Dist(ift.sqrt(pspec))
N = ift.ScalingOperator(noise, space) N = ift.ScalingOperator(space, noise)
n = N.draw_sample() n = N.draw_sample()
R = ift.ScalingOperator(10., space) R = ift.ScalingOperator(space, 10.)
def d_model(): def d_model():
if nonlinearity == "": if nonlinearity == "":
......
...@@ -58,7 +58,7 @@ def test_dataconv(): ...@@ -58,7 +58,7 @@ def test_dataconv():
def test_blockdiagonal(): def test_blockdiagonal():
op = ift.BlockDiagonalOperator( op = ift.BlockDiagonalOperator(
dom, {"d1": ift.ScalingOperator(20., dom["d1"])}) dom, {"d1": ift.ScalingOperator(dom["d1"], 20.)})
op2 = op(op) op2 = op(op)
ift.extra.consistency_check(op2) ift.extra.consistency_check(op2)
assert_equal(type(op2), ift.BlockDiagonalOperator) assert_equal(type(op2), ift.BlockDiagonalOperator)
......
...@@ -65,7 +65,7 @@ def test_times_inverse_times(space1, space2): ...@@ -65,7 +65,7 @@ def test_times_inverse_times(space1, space2):
def test_sum(space1): def test_sum(space1):
op1 = ift.makeOp(ift.Field.full(space1, 2.)) op1 = ift.makeOp(ift.Field.full(space1, 2.))
op2 = ift.ScalingOperator(3., space1) op2 = ift.ScalingOperator(space1, 3.)
full_op = op1 + op2 - (op2 - op1) + op1 + op1 + op2 full_op = op1 + op2 - (op2 - op1) + op1 + op1 + op2
x = ift.Field.full(space1, 1.) x = ift.Field.full(space1, 1.)
res = full_op(x) res = full_op(x)
...@@ -75,7 +75,7 @@ def test_sum(space1): ...@@ -75,7 +75,7 @@ def test_sum(space1):
def test_chain(space1): def test_chain(space1):
op1 = ift.makeOp(ift.Field.full(space1, 2.)) op1 = ift.makeOp(ift.Field.full(space1, 2.))
op2 = ift.ScalingOperator(3., space1) op2 = ift.ScalingOperator(space1, 3.,)
full_op = op1(op2)(op2)(op1)(op1)(op1)(op2) full_op = op1(op2)(op2)(op1)(op1)(op1)(op2)
x = ift.Field.full(space1, 1.) x = ift.Field.full(space1, 1.)
res = full_op(x) res = full_op(x)
...@@ -85,7 +85,7 @@ def test_chain(space1): ...@@ -85,7 +85,7 @@ def test_chain(space1):
def test_mix(space1): def test_mix(space1):
op1 = ift.makeOp(ift.Field.full(space1, 2.)) op1 = ift.makeOp(ift.Field.full(space1, 2.))
op2 = ift.ScalingOperator(3., space1) op2 = ift.ScalingOperator(space1, 3.)
full_op = op1(op2 + op2)(op1)(op1) - op1(op2) full_op = op1(op2 + op2)(op1)(op1) - op1(op2)
x = ift.Field.full(space1, 1.) x = ift.Field.full(space1, 1.)
res = full_op(x) res = full_op(x)
......
...@@ -39,10 +39,10 @@ seed = list2fixture([4, 78, 23]) ...@@ -39,10 +39,10 @@ seed = list2fixture([4, 78, 23])
def testBasics(space, seed): def testBasics(space, seed):
np.random.seed(seed) np.random.seed(seed)
S = ift.ScalingOperator(1., space) S = ift.ScalingOperator(space, 1.)
s = S.draw_sample() s = S.draw_sample()
var = ift.Linearization.make_var(s) var = ift.Linearization.make_var(s)
model = ift.ScalingOperator(6., var.target) model = ift.ScalingOperator(var.target, 6.)
ift.extra.check_jacobian_consistency(model, var.val) ift.extra.check_jacobian_consistency(model, var.val)
...@@ -64,7 +64,7 @@ def testBinary(type1, type2, space, seed): ...@@ -64,7 +64,7 @@ def testBinary(type1, type2, space, seed):
model = select_s1.scale(3.) model = select_s1.scale(3.)
pos = ift.from_random("normal", dom1) pos = ift.from_random("normal", dom1)
ift.extra.check_jacobian_consistency(model, pos, ntries=20) ift.extra.check_jacobian_consistency(model, pos, ntries=20)
model = ift.ScalingOperator(2.456, space)(select_s1*select_s2) model = ift.ScalingOperator(space, 2.456)(select_s1*select_s2)
pos = ift.from_random("normal", dom) pos = ift.from_random("normal", dom)
ift.extra.check_jacobian_consistency(model, pos, ntries=20) ift.extra.check_jacobian_consistency(model, pos, ntries=20)
model = ift.sigmoid(2.456*(select_s1*select_s2)) model = ift.sigmoid(2.456*(select_s1*select_s2))
...@@ -90,7 +90,7 @@ def testBinary(type1, type2, space, seed): ...@@ -90,7 +90,7 @@ def testBinary(type1, type2, space, seed):
def testPointModel(space, seed): def testPointModel(space, seed):
S = ift.ScalingOperator(1., space) S = ift.ScalingOperator(space, 1.)
pos = S.draw_sample() pos = S.draw_sample()
alpha = 1.5 alpha = 1.5
q = 0.73 q = 0.73
...@@ -118,7 +118,7 @@ def testDynamicModel(target, causal, minimum_phase, seed): ...@@ -118,7 +118,7 @@ def testDynamicModel(target, causal, minimum_phase, seed):
'minimum_phase': minimum_phase 'minimum_phase': minimum_phase
} }
model, _ = ift.dynamic_operator(**dct) model, _ = ift.dynamic_operator(**dct)
S = ift.ScalingOperator(1., model.domain) S = ift.ScalingOperator(model.domain, 1.)
pos = S.draw_sample() pos = S.draw_sample()
# FIXME I dont know why smaller tol fails for 3D example # FIXME I dont know why smaller tol fails for 3D example
ift.extra.check_jacobian_consistency(model, pos, tol=1e-5, ntries=20) ift.extra.check_jacobian_consistency(model, pos, tol=1e-5, ntries=20)
...@@ -139,7 +139,7 @@ def testDynamicModel(target, causal, minimum_phase, seed): ...@@ -139,7 +139,7 @@ def testDynamicModel(target, causal, minimum_phase, seed):
dct['sigc'] = 1. dct['sigc'] = 1.
dct['quant'] = 5 dct['quant'] = 5
model, _ = ift.dynamic_lightcone_operator(**dct) model, _ = ift.dynamic_lightcone_operator(**dct)
S = ift.ScalingOperator(1., model.domain) S = ift.ScalingOperator(model.domain, 1.)
pos = S.draw_sample() pos = S.draw_sample()
# FIXME I dont know why smaller tol fails for 3D example # FIXME I dont know why smaller tol fails for 3D example
ift.extra.check_jacobian_consistency( ift.extra.check_jacobian_consistency(
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment