descent_minimizers.py 15.2 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
# 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/>.
Theo Steininger's avatar
Theo Steininger committed
13
#
14
# Copyright(C) 2013-2019 Max-Planck-Society
Theo Steininger's avatar
Theo Steininger committed
15
#
16
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik.
17

18
import numpy as np
Philipp Arras's avatar
Philipp Arras committed
19

20
from ..logger import logger
21
22
from ..probing import approximation2endo
from ..sugar import makeOp
Philipp Arras's avatar
Philipp Arras committed
23
from .conjugate_gradient import ConjugateGradient
24
25
from .iteration_controllers import (AbsDeltaEnergyController,
                                    GradientNormController)
Martin Reinecke's avatar
Martin Reinecke committed
26
from .line_search import LineSearch
27
from .minimizer import Minimizer
Philipp Arras's avatar
Philipp Arras committed
28
from .quadratic_energy import QuadraticEnergy
29
30
31


class DescentMinimizer(Minimizer):
32
    """A base class used by gradient methods to find a local minimum.
33
34
35
36
37
38
39
40
41
42
43
44

    Descent minimization methods are used to find a local minimum of a scalar
    function by following a descent direction. This class implements the
    minimization procedure once a descent direction is known. The descent
    direction has to be implemented separately.

    Parameters
    ----------
    controller : IterationController
        Object that decides when to terminate the minimization.
    line_searcher : callable *optional*
        Function which infers the step size in the descent direction
Martin Reinecke's avatar
Martin Reinecke committed
45
        (default : LineSearch()).
46
47
    """

Martin Reinecke's avatar
Martin Reinecke committed
48
    def __init__(self, controller, line_searcher=LineSearch()):
49
50
51
        self._controller = controller
        self.line_searcher = line_searcher

52
    def __call__(self, energy):
53
        """Performs the minimization of the provided Energy functional.
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87

        Parameters
        ----------
        energy : Energy
           Energy object which provides value, gradient and metric at a
           specific position in parameter space.

        Returns
        -------
        Energy
            Latest `energy` of the minimization.
        int
            Can be controller.CONVERGED or controller.ERROR

        Notes
        -----
        The minimization is stopped if
            * the controller returns controller.CONVERGED or controller.ERROR,
            * a perfectly flat point is reached,
            * according to the line-search the minimum is found,
        """
        f_k_minus_1 = None
        controller = self._controller
        status = controller.start(energy)
        if status != controller.CONTINUE:
            return energy, status

        while True:
            # check if position is at a flat point
            if energy.gradient_norm == 0:
                return energy, controller.CONVERGED

            # compute a step length that reduces energy.value sufficiently
            new_energy, success = self.line_searcher.perform_line_search(
88
89
                energy=energy,
                pk=self.get_descent_direction(energy, f_k_minus_1),
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
                f_k_minus_1=f_k_minus_1)
            if not success:
                self.reset()

            f_k_minus_1 = energy.value

            if new_energy.value > energy.value:
                logger.error("Error: Energy has increased")
                return energy, controller.ERROR

            if new_energy.value == energy.value:
                logger.warning(
                    "Warning: Energy has not changed. Assuming convergence...")
                return new_energy, controller.CONVERGED

            energy = new_energy
            status = self._controller.check(energy)
            if status != controller.CONTINUE:
                return energy, status

    def reset(self):
        pass

113
    def get_descent_direction(self, energy, old_value=None):
114
        """Calculates the next descent direction.
115
116
117
118
119
120
121

        Parameters
        ----------
        energy : Energy
            An instance of the Energy class which shall be minimized. The
            position of `energy` is used as the starting point of minimization.

122
123
124
125
        old_value : float
            if provided, this must be the value of the energy in the previous
            step.

126
127
128
129
130
131
132
133
134
        Returns
        -------
        Field
           The descent direction.
        """
        raise NotImplementedError


class SteepestDescent(DescentMinimizer):
135
    """Implementation of the steepest descent minimization scheme.
136
137
138
139
140

    Also known as 'gradient descent'. This algorithm simply follows the
    functional's gradient for minimization.
    """

141
    def get_descent_direction(self, energy, _=None):
142
143
144
        return -energy.gradient


Martin Reinecke's avatar
Martin Reinecke committed
145
class RelaxedNewton(DescentMinimizer):
146
    """Calculates the descent direction according to a Newton scheme.
Martin Reinecke's avatar
Martin Reinecke committed
147
148
149
150
151
152
153

    The descent direction is determined by weighting the gradient at the
    current parameter position with the inverse local metric.
    """

    def __init__(self, controller, line_searcher=None):
        if line_searcher is None:
Martin Reinecke's avatar
Martin Reinecke committed
154
            line_searcher = LineSearch(preferred_initial_step_size=1.)
Martin Reinecke's avatar
Martin Reinecke committed
155
156
157
        super(RelaxedNewton, self).__init__(controller=controller,
                                            line_searcher=line_searcher)

158
    def get_descent_direction(self, energy, _=None):
Martin Reinecke's avatar
Martin Reinecke committed
159
160
161
        return -energy.metric.inverse_times(energy.gradient)


162
class NewtonCG(DescentMinimizer):
163
    """Calculates the descent direction according to a Newton-CG scheme.
164
165
166
167

    Algorithm derived from SciPy sources.
    """

168
    def __init__(self, controller, napprox=0, line_searcher=None, name=None,
Martin Reinecke's avatar
Martin Reinecke committed
169
170
                 nreset=20, max_cg_iterations=200, energy_reduction_factor=0.1,
                 file_name=None):
171
        if line_searcher is None:
Martin Reinecke's avatar
Martin Reinecke committed
172
            line_searcher = LineSearch(preferred_initial_step_size=1.)
173
174
        super(NewtonCG, self).__init__(controller=controller,
                                       line_searcher=line_searcher)
175
        self._napprox = napprox
176
        self._name = name
177
        self._nreset = nreset
178
        self._file_name = file_name
179
180
        self._max_cg_iterations = max_cg_iterations
        self._alpha = energy_reduction_factor
181

182
183
    def get_descent_direction(self, energy, old_value=None):
        if old_value is None:
184
            ic = GradientNormController(iteration_limit=5)
185
        else:
186
            ediff = self._alpha*(old_value-energy.value)
187
            ic = AbsDeltaEnergyController(
Martin Reinecke's avatar
Martin Reinecke committed
188
                ediff, iteration_limit=self._max_cg_iterations, name=self._name,
189
                file_name=self._file_name)
190
        e = QuadraticEnergy(0*energy.position, energy.metric, energy.gradient)
191
192
        p = None
        if self._napprox > 1:
Reimar H Leike's avatar
fixup    
Reimar H Leike committed
193
194
            met = energy.metric
            p = makeOp(approximation2endo(met, self._napprox)).inverse
Philipp Arras's avatar
Philipp Arras committed
195
        e, conv = ConjugateGradient(ic, nreset=self._nreset)(e, p)
Philipp Arras's avatar
Philipp Arras committed
196
        return -e.position
197
198


199
class L_BFGS(DescentMinimizer):
Martin Reinecke's avatar
Martin Reinecke committed
200
    def __init__(self, controller, line_searcher=LineSearch(),
201
202
203
204
205
206
207
208
209
210
211
212
213
214
                 max_history_length=5):
        super(L_BFGS, self).__init__(controller=controller,
                                     line_searcher=line_searcher)
        self.max_history_length = max_history_length

    def __call__(self, energy):
        self.reset()
        return super(L_BFGS, self).__call__(energy)

    def reset(self):
        self._k = 0
        self._s = [None]*self.max_history_length
        self._y = [None]*self.max_history_length

215
    def get_descent_direction(self, energy, _=None):
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
        x = energy.position
        s = self._s
        y = self._y
        k = self._k
        maxhist = self.max_history_length
        gradient = energy.gradient

        nhist = min(k, maxhist)
        alpha = [None]*maxhist
        p = -gradient
        if k > 0:
            idx = (k-1) % maxhist
            s[idx] = x-self._lastx
            y[idx] = gradient-self._lastgrad
        if nhist > 0:
            for i in range(k-1, k-nhist-1, -1):
                idx = i % maxhist
                alpha[idx] = s[idx].vdot(p)/s[idx].vdot(y[idx])
                p = p - alpha[idx]*y[idx]
            idx = (k-1) % maxhist
            fact = s[idx].vdot(y[idx]) / y[idx].vdot(y[idx])
            if fact <= 0.:
                logger.error("L-BFGS curvature not positive definite!")
            p = p*fact
            for i in range(k-nhist, k):
                idx = i % maxhist
                beta = y[idx].vdot(p) / s[idx].vdot(y[idx])
                p = p + (alpha[idx]-beta)*s[idx]
        self._lastx = x
        self._lastgrad = gradient
        self._k += 1
        return p
theos's avatar
theos committed
248
249


250
class VL_BFGS(DescentMinimizer):
Martin Reinecke's avatar
Martin Reinecke committed
251
252
253
254
255
256
257
258
259
260
261
262
263
264
    """Implementation of the Vector-free L-BFGS minimization scheme.

    Find the descent direction by using the inverse Hessian.
    Instead of storing the whole matrix, it stores only the last few
    updates, which are used to do operations requiring the inverse
    Hessian product. The updates are represented in a new basis to optimize
    the algorithm.

    References
    ----------
    W. Chen, Z. Wang, J. Zhou, "Large-scale L-BFGS using MapReduce", 2014,
    Microsoft
    """

Martin Reinecke's avatar
Martin Reinecke committed
265
    def __init__(self, controller, line_searcher=LineSearch(),
Martin Reinecke's avatar
Martin Reinecke committed
266
                 max_history_length=5):
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
267
268
        super(VL_BFGS, self).__init__(controller=controller,
                                      line_searcher=line_searcher)
269
270
        self.max_history_length = max_history_length

271
    def __call__(self, energy):
272
        self._information_store = None
273
        return super(VL_BFGS, self).__call__(energy)
274

Martin Reinecke's avatar
stage 1    
Martin Reinecke committed
275
276
277
    def reset(self):
        self._information_store = None

278
    def get_descent_direction(self, energy, _=None):
279
280
        x = energy.position
        gradient = energy.gradient
281
282
283
284
        # initialize the information store if it doesn't already exist
        try:
            self._information_store.add_new_point(x, gradient)
        except AttributeError:
Martin Reinecke's avatar
Martin Reinecke committed
285
286
            self._information_store = _InformationStore(
                self.max_history_length, x0=x, gradient=gradient)
287
288
289
290

        b = self._information_store.b
        delta = self._information_store.delta

291
        descent_direction = delta[0] * b[0]
Martin Reinecke's avatar
Martin Reinecke committed
292
        for i in range(1, len(delta)):
293
            descent_direction = descent_direction + delta[i]*b[i]
294

295
        return descent_direction
theos's avatar
theos committed
296
297


Martin Reinecke's avatar
Martin Reinecke committed
298
class _InformationStore(object):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
299
    """Class for storing a list of past updates.
300

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
301
302
    Parameters
    ----------
Martin Reinecke's avatar
Martin Reinecke committed
303
    max_history_length : int
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
304
305
306
307
308
        Maximum number of stored past updates.
    x0 : Field
        Initial position in variable space.
    gradient : Field
        Gradient at position x0.
309

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
310
311
    Attributes
    ----------
Martin Reinecke's avatar
Martin Reinecke committed
312
    max_history_length : int
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
313
314
        Maximum number of stored past updates.
    s : List
Martin Reinecke's avatar
Martin Reinecke committed
315
        Circular buffer of past position differences, which are Fields.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
316
    y : List
Martin Reinecke's avatar
Martin Reinecke committed
317
        Circular buffer of past gradient differences, which are Fields.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
318
    last_x : Field
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
319
        Latest position in variable space.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
320
    last_gradient : Field
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
321
        Gradient at latest position.
Martin Reinecke's avatar
Martin Reinecke committed
322
    k : int
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
323
        Number of updates that have taken place
Martin Reinecke's avatar
Martin Reinecke committed
324
    ss : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
325
        2D circular buffer of scalar products between different elements of s.
Martin Reinecke's avatar
Martin Reinecke committed
326
    sy : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
327
        2D circular buffer of scalar products between elements of s and y.
Martin Reinecke's avatar
Martin Reinecke committed
328
    yy : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
329
        2D circular buffer of scalar products between different elements of y.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
330
    """
Philipp Arras's avatar
Philipp Arras committed
331

332
333
    def __init__(self, max_history_length, x0, gradient):
        self.max_history_length = max_history_length
334
335
        self.s = [None]*max_history_length
        self.y = [None]*max_history_length
336
337
        self.last_x = x0
        self.last_gradient = gradient
theos's avatar
theos committed
338
        self.k = 0
339

Martin Reinecke's avatar
Martin Reinecke committed
340
        mmax = max_history_length
Martin Reinecke's avatar
Martin Reinecke committed
341
342
343
        self.ss = np.empty((mmax, mmax), dtype=np.float64)
        self.sy = np.empty((mmax, mmax), dtype=np.float64)
        self.yy = np.empty((mmax, mmax), dtype=np.float64)
344
345
346

    @property
    def history_length(self):
Martin Reinecke's avatar
Martin Reinecke committed
347
        """Returns the number of currently stored updates."""
348
349
350
351
        return min(self.k, self.max_history_length)

    @property
    def b(self):
352
353
        """Combines s, y and gradient to form the new base vectors b.

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
354
355
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
356
        List
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
357
358
            List of new basis vectors.
        """
359
360
        result = []
        m = self.history_length
Martin Reinecke's avatar
Martin Reinecke committed
361
        mmax = self.max_history_length
362

Martin Reinecke's avatar
Martin Reinecke committed
363
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
364
            result.append(self.s[(self.k-m+i) % mmax])
365

Martin Reinecke's avatar
Martin Reinecke committed
366
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
367
            result.append(self.y[(self.k-m+i) % mmax])
368
369
370
371
372
373
374

        result.append(self.last_gradient)

        return result

    @property
    def b_dot_b(self):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
375
        """Generates the (2m+1) * (2m+1) scalar matrix.
376

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
377
        The i,j-th element of the matrix is a scalar product between the i-th
378
379
        and j-th base vector.

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
380
381
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
382
        numpy.ndarray
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
383
384
            Scalar matrix.
        """
385
        m = self.history_length
Martin Reinecke's avatar
Martin Reinecke committed
386
        mmax = self.max_history_length
387
388
389
        k = self.k
        result = np.empty((2*m+1, 2*m+1), dtype=np.float)

Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
390
        # update the stores
Martin Reinecke's avatar
Martin Reinecke committed
391
        k1 = (k-1) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
392
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
393
            kmi = (k-m+i) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
394
395
396
            self.ss[kmi, k1] = self.ss[k1, kmi] = self.s[kmi].vdot(self.s[k1])
            self.yy[kmi, k1] = self.yy[k1, kmi] = self.y[kmi].vdot(self.y[k1])
            self.sy[kmi, k1] = self.s[kmi].vdot(self.y[k1])
Martin Reinecke's avatar
Martin Reinecke committed
397
        for j in range(m-1):
Martin Reinecke's avatar
Martin Reinecke committed
398
399
            kmj = (k-m+j) % mmax
            self.sy[k1, kmj] = self.s[k1].vdot(self.y[kmj])
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
400

Martin Reinecke's avatar
Martin Reinecke committed
401
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
402
            kmi = (k-m+i) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
403
            for j in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
404
                kmj = (k-m+j) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
405
406
407
                result[i, j] = self.ss[kmi, kmj]
                result[i, m+j] = result[m+j, i] = self.sy[kmi, kmj]
                result[m+i, m+j] = self.yy[kmi, kmj]
408

409
            sgrad_i = self.s[kmi].vdot(self.last_gradient)
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
410
            result[2*m, i] = result[i, 2*m] = sgrad_i
411

Martin Reinecke's avatar
fix    
Martin Reinecke committed
412
            ygrad_i = self.y[kmi].vdot(self.last_gradient)
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
413
            result[2*m, m+i] = result[m+i, 2*m] = ygrad_i
414

Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
415
        result[2*m, 2*m] = self.last_gradient.norm()
416
        return result
theos's avatar
theos committed
417
418

    @property
419
    def delta(self):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
420
        """Calculates the new scalar coefficients (deltas).
421

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
422
423
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
424
        List
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
425
426
            List of the new scalar coefficients (deltas).
        """
427
428
429
430
431
432
433
434
        m = self.history_length
        b_dot_b = self.b_dot_b

        delta = np.zeros(2*m+1, dtype=np.float)
        delta[2*m] = -1

        alpha = np.empty(m, dtype=np.float)

Martin Reinecke's avatar
Martin Reinecke committed
435
436
        for j in range(m-1, -1, -1):
            delta_b_b = sum([delta[l] * b_dot_b[l, j] for l in range(2*m+1)])
437
438
439
            alpha[j] = delta_b_b/b_dot_b[j, m+j]
            delta[m+j] -= alpha[j]

Martin Reinecke's avatar
Martin Reinecke committed
440
        for i in range(2*m+1):
441
442
            delta[i] *= b_dot_b[m-1, 2*m-1]/b_dot_b[2*m-1, 2*m-1]

Martin Reinecke's avatar
Martin Reinecke committed
443
        for j in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
444
            delta_b_b = sum([delta[l]*b_dot_b[m+j, l] for l in range(2*m+1)])
445
446
447
448
449
            beta = delta_b_b/b_dot_b[j, m+j]
            delta[j] += (alpha[j] - beta)

        return delta

theos's avatar
theos committed
450
    def add_new_point(self, x, gradient):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
451
452
        """Updates the s list and y list.

Martin Reinecke's avatar
Martin Reinecke committed
453
454
        Calculates the new position and gradient differences and enters them
        into the respective list.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
455
        """
Martin Reinecke's avatar
Martin Reinecke committed
456
457
458
        mmax = self.max_history_length
        self.s[self.k % mmax] = x - self.last_x
        self.y[self.k % mmax] = gradient - self.last_gradient
theos's avatar
theos committed
459

460
461
        self.last_x = x
        self.last_gradient = gradient
theos's avatar
theos committed
462

Martin Reinecke's avatar
fixes    
Martin Reinecke committed
463
        self.k += 1