operator.py 8.25 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
#
# Copyright(C) 2013-2019 Max-Planck-Society
#
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik.
Martin Reinecke's avatar
Martin Reinecke committed
17

Martin Reinecke's avatar
Martin Reinecke committed
18
import numpy as np
Philipp Arras's avatar
Philipp Arras committed
19
from ..utilities import NiftyMetaBase, indent
Martin Reinecke's avatar
Martin Reinecke committed
20 21 22 23 24 25 26


class Operator(NiftyMetaBase()):
    """Transforms values living on one domain into values living on another
    domain, and can also provide the Jacobian.
    """

Martin Reinecke's avatar
Martin Reinecke committed
27
    @property
Martin Reinecke's avatar
Martin Reinecke committed
28 29 30 31
    def domain(self):
        """DomainTuple or MultiDomain : the operator's input domain

            The domain on which the Operator's input Field lives."""
Martin Reinecke's avatar
Martin Reinecke committed
32
        return self._domain
Martin Reinecke's avatar
Martin Reinecke committed
33

Martin Reinecke's avatar
Martin Reinecke committed
34
    @property
Martin Reinecke's avatar
Martin Reinecke committed
35 36 37 38
    def target(self):
        """DomainTuple or MultiDomain : the operator's output domain

            The domain on which the Operator's output Field lives."""
Martin Reinecke's avatar
Martin Reinecke committed
39
        return self._target
Martin Reinecke's avatar
Martin Reinecke committed
40

Martin Reinecke's avatar
Martin Reinecke committed
41 42 43 44 45 46
    @staticmethod
    def _check_domain_equality(dom_op, dom_field):
        if dom_op != dom_field:
            s = "The operator's and field's domains don't match."
            from ..domain_tuple import DomainTuple
            from ..multi_domain import MultiDomain
Sebastian Hutschenreuter's avatar
fix  
Sebastian Hutschenreuter committed
47
            if not isinstance(dom_op, (DomainTuple, MultiDomain,)):
Martin Reinecke's avatar
Martin Reinecke committed
48 49 50 51
                s += " Your operator's domain is neither a `DomainTuple`" \
                     " nor a `MultiDomain`."
            raise ValueError(s)

Martin Reinecke's avatar
cleanup  
Martin Reinecke committed
52 53 54 55 56 57 58 59 60 61
    def scale(self, factor):
        if factor == 1:
            return self
        from .scaling_operator import ScalingOperator
        return ScalingOperator(factor, self.target)(self)

    def conjugate(self):
        from .simple_linear_operators import ConjugationOperator
        return ConjugationOperator(self.target)(self)

Martin Reinecke's avatar
Martin Reinecke committed
62 63 64 65 66
    @property
    def real(self):
        from .simple_linear_operators import Realizer
        return Realizer(self.target)(self)

Martin Reinecke's avatar
cleanup  
Martin Reinecke committed
67 68 69
    def __neg__(self):
        return self.scale(-1)

Martin Reinecke's avatar
Martin Reinecke committed
70 71 72
    def __matmul__(self, x):
        if not isinstance(x, Operator):
            return NotImplemented
Martin Reinecke's avatar
Martin Reinecke committed
73
        return _OpChain.make((self, x))
Martin Reinecke's avatar
Martin Reinecke committed
74

Martin Reinecke's avatar
Martin Reinecke committed
75
    def __mul__(self, x):
76 77 78 79 80
        if isinstance(x, Operator):
            return _OpProd(self, x)
        if np.isscalar(x):
            return self.scale(x)
        return NotImplemented
Martin Reinecke's avatar
Martin Reinecke committed
81

Philipp Arras's avatar
Philipp Arras committed
82 83 84
    def __add__(self, x):
        if not isinstance(x, Operator):
            return NotImplemented
Martin Reinecke's avatar
Martin Reinecke committed
85
        return _OpSum(self, x)
Philipp Arras's avatar
Philipp Arras committed
86

87 88 89 90 91
    def __sub__(self, x):
        if not isinstance(x, Operator):
            return NotImplemented
        return _OpSum(self, -x)

Martin Reinecke's avatar
Martin Reinecke committed
92 93 94 95 96
    def __pow__(self, power):
        if not np.isscalar(power):
            return NotImplemented
        return _OpChain.make((_PowerOp(self.target, power), self))

Martin Reinecke's avatar
Martin Reinecke committed
97 98
    def apply(self, x):
        raise NotImplementedError
Martin Reinecke's avatar
Martin Reinecke committed
99

Philipp Arras's avatar
Philipp Arras committed
100
    def force(self, x):
Philipp Arras's avatar
Philipp Arras committed
101
        """Extract correct subset of domain of x and apply operator."""
Philipp Arras's avatar
Philipp Arras committed
102 103
        return self.apply(x.extract(self.domain))

104 105 106
    def _check_input(self, x):
        from ..linearization import Linearization
        d = x.target if isinstance(x, Linearization) else x.domain
Martin Reinecke's avatar
Martin Reinecke committed
107
        self._check_domain_equality(self._domain, d)
108

Martin Reinecke's avatar
Martin Reinecke committed
109
    def __call__(self, x):
Martin Reinecke's avatar
Martin Reinecke committed
110 111 112
        if isinstance(x, Operator):
            return _OpChain.make((self, x))
        return self.apply(x)
Martin Reinecke's avatar
Martin Reinecke committed
113

Martin Reinecke's avatar
Martin Reinecke committed
114 115 116 117 118 119 120 121
    def ducktape(self, name):
        from .simple_linear_operators import ducktape
        return self(ducktape(self, None, name))

    def ducktape_left(self, name):
        from .simple_linear_operators import ducktape
        return ducktape(None, self, name)(self)

Martin Reinecke's avatar
Martin Reinecke committed
122 123 124
    def __repr__(self):
        return self.__class__.__name__

Martin Reinecke's avatar
Martin Reinecke committed
125

Philipp Arras's avatar
Philipp Arras committed
126
for f in ["sqrt", "exp", "log", "tanh", "positive_tanh", 'clipped_exp']:
Martin Reinecke's avatar
Martin Reinecke committed
127 128
    def func(f):
        def func2(self):
129
            fa = _FunctionApplier(self.target, f)
Martin Reinecke's avatar
Martin Reinecke committed
130 131 132 133 134 135 136 137
            return _OpChain.make((fa, self))
        return func2
    setattr(Operator, f, func(f))


class _FunctionApplier(Operator):
    def __init__(self, domain, funcname):
        from ..sugar import makeDomain
Martin Reinecke's avatar
Martin Reinecke committed
138
        self._domain = self._target = makeDomain(domain)
Martin Reinecke's avatar
Martin Reinecke committed
139 140
        self._funcname = funcname

Martin Reinecke's avatar
Martin Reinecke committed
141
    def apply(self, x):
142
        self._check_input(x)
Martin Reinecke's avatar
Martin Reinecke committed
143 144 145
        return getattr(x, self._funcname)()


Martin Reinecke's avatar
Martin Reinecke committed
146 147 148 149 150 151 152 153 154 155 156
class _PowerOp(Operator):
    def __init__(self, domain, power):
        from ..sugar import makeDomain
        self._domain = self._target = makeDomain(domain)
        self._power = power

    def apply(self, x):
        self._check_input(x)
        return x**self._power


Martin Reinecke's avatar
Martin Reinecke committed
157 158 159 160 161 162 163 164 165 166
class _CombinedOperator(Operator):
    def __init__(self, ops, _callingfrommake=False):
        if not _callingfrommake:
            raise NotImplementedError
        self._ops = tuple(ops)

    @classmethod
    def unpack(cls, ops, res):
        for op in ops:
            if isinstance(op, cls):
Martin Reinecke's avatar
Martin Reinecke committed
167
                res = cls.unpack(op._ops, res)
Martin Reinecke's avatar
Martin Reinecke committed
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
            else:
                res = res + [op]
        return res

    @classmethod
    def make(cls, ops):
        res = cls.unpack(ops, [])
        if len(res) == 1:
            return res[0]
        return cls(res, _callingfrommake=True)


class _OpChain(_CombinedOperator):
    def __init__(self, ops, _callingfrommake=False):
        super(_OpChain, self).__init__(ops, _callingfrommake)
Martin Reinecke's avatar
Martin Reinecke committed
183 184
        self._domain = self._ops[-1].domain
        self._target = self._ops[0].target
Martin Reinecke's avatar
Martin Reinecke committed
185 186 187
        for i in range(1, len(self._ops)):
            if self._ops[i-1].domain != self._ops[i].target:
                raise ValueError("domain mismatch")
Martin Reinecke's avatar
Martin Reinecke committed
188

Martin Reinecke's avatar
Martin Reinecke committed
189
    def apply(self, x):
190
        self._check_input(x)
Martin Reinecke's avatar
Martin Reinecke committed
191 192 193 194 195
        for op in reversed(self._ops):
            x = op(x)
        return x


Philipp Arras's avatar
Philipp Arras committed
196 197 198 199 200
    def __repr__(self):
        subs = "\n".join(sub.__repr__() for sub in self._ops)
        return "_OpChain:\n" + indent(subs)


Martin Reinecke's avatar
Martin Reinecke committed
201 202 203 204 205 206 207 208 209
class _OpProd(Operator):
    def __init__(self, op1, op2):
        from ..sugar import domain_union
        self._domain = domain_union((op1.domain, op2.domain))
        self._target = op1.target
        if op1.target != op2.target:
            raise ValueError("target mismatch")
        self._op1 = op1
        self._op2 = op2
Martin Reinecke's avatar
Martin Reinecke committed
210

Martin Reinecke's avatar
Martin Reinecke committed
211
    def apply(self, x):
Martin Reinecke's avatar
Martin Reinecke committed
212 213
        from ..linearization import Linearization
        from ..sugar import makeOp
214
        self._check_input(x)
Martin Reinecke's avatar
Martin Reinecke committed
215
        lin = isinstance(x, Linearization)
216 217 218
        v = x._val if lin else x
        v1 = v.extract(self._op1.domain)
        v2 = v.extract(self._op2.domain)
Martin Reinecke's avatar
Martin Reinecke committed
219
        if not lin:
220
            return self._op1(v1) * self._op2(v2)
221 222 223
        wm = x.want_metric
        lin1 = self._op1(Linearization.make_var(v1, wm))
        lin2 = self._op2(Linearization.make_var(v2, wm))
Martin Reinecke's avatar
Martin Reinecke committed
224 225
        op = (makeOp(lin1._val)(lin2._jac))._myadd(
            makeOp(lin2._val)(lin1._jac), False)
226
        return lin1.new(lin1._val*lin2._val, op(x.jac))
Martin Reinecke's avatar
Martin Reinecke committed
227 228


Philipp Arras's avatar
Philipp Arras committed
229 230 231 232 233
    def __repr__(self):
        subs = "\n".join(sub.__repr__() for sub in (self._op1, self._op2))
        return "_OpProd:\n"+indent(subs)


Martin Reinecke's avatar
Martin Reinecke committed
234 235
class _OpSum(Operator):
    def __init__(self, op1, op2):
Philipp Arras's avatar
Philipp Arras committed
236
        from ..sugar import domain_union
Martin Reinecke's avatar
Martin Reinecke committed
237 238 239 240
        self._domain = domain_union((op1.domain, op2.domain))
        self._target = domain_union((op1.target, op2.target))
        self._op1 = op1
        self._op2 = op2
Philipp Arras's avatar
Philipp Arras committed
241 242

    def apply(self, x):
Martin Reinecke's avatar
Martin Reinecke committed
243
        from ..linearization import Linearization
244
        self._check_input(x)
Martin Reinecke's avatar
Martin Reinecke committed
245 246 247 248
        lin = isinstance(x, Linearization)
        v = x._val if lin else x
        v1 = v.extract(self._op1.domain)
        v2 = v.extract(self._op2.domain)
Philipp Arras's avatar
Philipp Arras committed
249
        res = None
Martin Reinecke's avatar
Martin Reinecke committed
250 251
        if not lin:
            return self._op1(v1).unite(self._op2(v2))
252 253 254
        wm = x.want_metric
        lin1 = self._op1(Linearization.make_var(v1, wm))
        lin2 = self._op2(Linearization.make_var(v2, wm))
Martin Reinecke's avatar
Martin Reinecke committed
255
        op = lin1._jac._myadd(lin2._jac, False)
Martin Reinecke's avatar
bug fix  
Martin Reinecke committed
256
        res = lin1.new(lin1._val.unite(lin2._val), op(x.jac))
Martin Reinecke's avatar
Martin Reinecke committed
257 258
        if lin1._metric is not None and lin2._metric is not None:
            res = res.add_metric(lin1._metric + lin2._metric)
Philipp Arras's avatar
Philipp Arras committed
259
        return res
Philipp Arras's avatar
Philipp Arras committed
260 261 262 263

    def __repr__(self):
        subs = "\n".join(sub.__repr__() for sub in (self._op1, self._op2))
        return "_OpSum:\n"+indent(subs)