descent_minimizers.py 15 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
Martin Reinecke's avatar
Martin Reinecke committed
21
22
23
24
25
from ..probing import approximation2endo
from ..sugar import makeOp
from .conjugate_gradient import ConjugateGradient
from .iteration_controllers import (AbsDeltaEnergyController,
                                    GradientNormController)
Martin Reinecke's avatar
Martin Reinecke committed
26
from .line_search import LineSearch
27
from .minimizer import Minimizer
Martin Reinecke's avatar
Martin Reinecke 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
52
        self._controller = controller
        self.line_searcher = line_searcher

    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(
Martin Reinecke's avatar
Martin Reinecke committed
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

Martin Reinecke's avatar
Martin Reinecke committed
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.

Martin Reinecke's avatar
Martin Reinecke committed
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.
    """

Martin Reinecke's avatar
Martin Reinecke committed
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)

Martin Reinecke's avatar
Martin Reinecke committed
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.
    """

Martin Reinecke's avatar
Martin Reinecke committed
168
169
    def __init__(self, controller, napprox=0, line_searcher=None, name=None,
                 nreset=20):
170
        if line_searcher is None:
Martin Reinecke's avatar
Martin Reinecke committed
171
            line_searcher = LineSearch(preferred_initial_step_size=1.)
172
173
        super(NewtonCG, self).__init__(controller=controller,
                                       line_searcher=line_searcher)
Martin Reinecke's avatar
Martin Reinecke committed
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
        self._napprox = napprox
        self._name = name
        self._nreset = nreset

    def get_descent_direction(self, energy, old_value=None):
        if old_value is None:
            ic = GradientNormController(iteration_limit=5)
        else:
            alpha = 0.1
            ediff = alpha*(old_value-energy.value)
            ic = AbsDeltaEnergyController(
                ediff, iteration_limit=200, name=self._name)
        e = QuadraticEnergy(0*energy.position, energy.metric, energy.gradient)
        p = None
        if self._napprox > 1:
            unscmet, sc = energy.unscaled_metric()
            p = makeOp(approximation2endo(unscmet, self._napprox)*sc).inverse
        e, conv = ConjugateGradient(ic, nreset=self._nreset)(e, p)
        return -e.position
193
194


195
class L_BFGS(DescentMinimizer):
Martin Reinecke's avatar
Martin Reinecke committed
196
    def __init__(self, controller, line_searcher=LineSearch(),
197
198
199
200
201
202
203
204
205
206
207
208
209
210
                 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

Martin Reinecke's avatar
Martin Reinecke committed
211
    def get_descent_direction(self, energy, _=None):
212
213
214
215
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
        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
Theo Steininger's avatar
Theo Steininger committed
244
245


246
class VL_BFGS(DescentMinimizer):
Martin Reinecke's avatar
Martin Reinecke committed
247
248
249
250
251
252
253
254
255
256
257
258
259
260
    """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
261
    def __init__(self, controller, line_searcher=LineSearch(),
Martin Reinecke's avatar
Martin Reinecke committed
262
                 max_history_length=5):
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
263
264
        super(VL_BFGS, self).__init__(controller=controller,
                                      line_searcher=line_searcher)
265
266
        self.max_history_length = max_history_length

267
    def __call__(self, energy):
268
        self._information_store = None
269
        return super(VL_BFGS, self).__call__(energy)
270

Martin Reinecke's avatar
stage 1    
Martin Reinecke committed
271
272
273
    def reset(self):
        self._information_store = None

Martin Reinecke's avatar
Martin Reinecke committed
274
    def get_descent_direction(self, energy, _=None):
275
276
        x = energy.position
        gradient = energy.gradient
277
278
279
280
        # 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
281
282
            self._information_store = _InformationStore(
                self.max_history_length, x0=x, gradient=gradient)
283
284
285
286

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

287
        descent_direction = delta[0] * b[0]
Martin Reinecke's avatar
Martin Reinecke committed
288
        for i in range(1, len(delta)):
289
            descent_direction = descent_direction + delta[i]*b[i]
290

291
        return descent_direction
Theo Steininger's avatar
Theo Steininger committed
292
293


Martin Reinecke's avatar
Martin Reinecke committed
294
class _InformationStore(object):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
295
    """Class for storing a list of past updates.
296

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
297
298
    Parameters
    ----------
Martin Reinecke's avatar
Martin Reinecke committed
299
    max_history_length : int
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
300
301
302
303
304
        Maximum number of stored past updates.
    x0 : Field
        Initial position in variable space.
    gradient : Field
        Gradient at position x0.
305

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
306
307
    Attributes
    ----------
Martin Reinecke's avatar
Martin Reinecke committed
308
    max_history_length : int
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
309
310
        Maximum number of stored past updates.
    s : List
Martin Reinecke's avatar
Martin Reinecke committed
311
        Circular buffer of past position differences, which are Fields.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
312
    y : List
Martin Reinecke's avatar
Martin Reinecke committed
313
        Circular buffer of past gradient differences, which are Fields.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
314
    last_x : Field
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
315
        Latest position in variable space.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
316
    last_gradient : Field
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
317
        Gradient at latest position.
Martin Reinecke's avatar
Martin Reinecke committed
318
    k : int
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
319
        Number of updates that have taken place
Martin Reinecke's avatar
Martin Reinecke committed
320
    ss : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
321
        2D circular buffer of scalar products between different elements of s.
Martin Reinecke's avatar
Martin Reinecke committed
322
    sy : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
323
        2D circular buffer of scalar products between elements of s and y.
Martin Reinecke's avatar
Martin Reinecke committed
324
    yy : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
325
        2D circular buffer of scalar products between different elements of y.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
326
    """
Philipp Arras's avatar
Philipp Arras committed
327

328
329
    def __init__(self, max_history_length, x0, gradient):
        self.max_history_length = max_history_length
330
331
        self.s = [None]*max_history_length
        self.y = [None]*max_history_length
332
333
        self.last_x = x0
        self.last_gradient = gradient
Theo Steininger's avatar
Theo Steininger committed
334
        self.k = 0
335

Martin Reinecke's avatar
Martin Reinecke committed
336
        mmax = max_history_length
Martin Reinecke's avatar
Martin Reinecke committed
337
338
339
        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)
340
341
342

    @property
    def history_length(self):
Martin Reinecke's avatar
Martin Reinecke committed
343
        """Returns the number of currently stored updates."""
344
345
346
347
        return min(self.k, self.max_history_length)

    @property
    def b(self):
348
349
        """Combines s, y and gradient to form the new base vectors b.

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
350
351
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
352
        List
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
353
354
            List of new basis vectors.
        """
355
356
        result = []
        m = self.history_length
Martin Reinecke's avatar
Martin Reinecke committed
357
        mmax = self.max_history_length
358

Martin Reinecke's avatar
Martin Reinecke committed
359
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
360
            result.append(self.s[(self.k-m+i) % mmax])
361

Martin Reinecke's avatar
Martin Reinecke committed
362
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
363
            result.append(self.y[(self.k-m+i) % mmax])
364
365
366
367
368
369
370

        result.append(self.last_gradient)

        return result

    @property
    def b_dot_b(self):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
371
        """Generates the (2m+1) * (2m+1) scalar matrix.
372

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
373
        The i,j-th element of the matrix is a scalar product between the i-th
374
375
        and j-th base vector.

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
376
377
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
378
        numpy.ndarray
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
379
380
            Scalar matrix.
        """
381
        m = self.history_length
Martin Reinecke's avatar
Martin Reinecke committed
382
        mmax = self.max_history_length
383
384
385
        k = self.k
        result = np.empty((2*m+1, 2*m+1), dtype=np.float)

Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
386
        # update the stores
Martin Reinecke's avatar
Martin Reinecke committed
387
        k1 = (k-1) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
388
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
389
            kmi = (k-m+i) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
390
391
392
            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
393
        for j in range(m-1):
Martin Reinecke's avatar
Martin Reinecke committed
394
395
            kmj = (k-m+j) % mmax
            self.sy[k1, kmj] = self.s[k1].vdot(self.y[kmj])
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
396

Martin Reinecke's avatar
Martin Reinecke committed
397
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
398
            kmi = (k-m+i) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
399
            for j in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
400
                kmj = (k-m+j) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
401
402
403
                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]
404

405
            sgrad_i = self.s[kmi].vdot(self.last_gradient)
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
406
            result[2*m, i] = result[i, 2*m] = sgrad_i
407

Martin Reinecke's avatar
fix    
Martin Reinecke committed
408
            ygrad_i = self.y[kmi].vdot(self.last_gradient)
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
409
            result[2*m, m+i] = result[m+i, 2*m] = ygrad_i
410

Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
411
        result[2*m, 2*m] = self.last_gradient.norm()
412
        return result
Theo Steininger's avatar
Theo Steininger committed
413
414

    @property
415
    def delta(self):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
416
        """Calculates the new scalar coefficients (deltas).
417

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
418
419
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
420
        List
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
421
422
            List of the new scalar coefficients (deltas).
        """
423
424
425
426
427
428
429
430
        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
431
432
        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)])
433
434
435
            alpha[j] = delta_b_b/b_dot_b[j, m+j]
            delta[m+j] -= alpha[j]

Martin Reinecke's avatar
Martin Reinecke committed
436
        for i in range(2*m+1):
437
438
            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
439
        for j in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
440
            delta_b_b = sum([delta[l]*b_dot_b[m+j, l] for l in range(2*m+1)])
441
442
443
444
445
            beta = delta_b_b/b_dot_b[j, m+j]
            delta[j] += (alpha[j] - beta)

        return delta

Theo Steininger's avatar
Theo Steininger committed
446
    def add_new_point(self, x, gradient):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
447
448
        """Updates the s list and y list.

Martin Reinecke's avatar
Martin Reinecke committed
449
450
        Calculates the new position and gradient differences and enters them
        into the respective list.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
451
        """
Martin Reinecke's avatar
Martin Reinecke committed
452
453
454
        mmax = self.max_history_length
        self.s[self.k % mmax] = x - self.last_x
        self.y[self.k % mmax] = gradient - self.last_gradient
Theo Steininger's avatar
Theo Steininger committed
455

456
457
        self.last_x = x
        self.last_gradient = gradient
Theo Steininger's avatar
Theo Steininger committed
458

Martin Reinecke's avatar
fixes    
Martin Reinecke committed
459
        self.k += 1