descent_minimizers.py 13.5 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
#
Martin Reinecke's avatar
Martin Reinecke committed
14
# Copyright(C) 2013-2018 Max-Planck-Society
Theo Steininger's avatar
Theo Steininger committed
15
16
17
#
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik
# and financially supported by the Studienstiftung des deutschen Volkes.
theos's avatar
theos committed
18

19
from __future__ import absolute_import, division, print_function
20

21
import numpy as np
22
from ..compat import *
23
from ..logger import logger
Martin Reinecke's avatar
Martin Reinecke committed
24
from .line_search_strong_wolfe import LineSearchStrongWolfe
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
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
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
from .minimizer import Minimizer


class DescentMinimizer(Minimizer):
    """ A base class used by gradient methods to find a local minimum.

    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
        (default : LineSearchStrongWolfe()).
    """

    def __init__(self, controller, line_searcher=LineSearchStrongWolfe()):
        self._controller = controller
        self.line_searcher = line_searcher

    def __call__(self, energy):
        """ Performs the minimization of the provided Energy functional.

        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(
                energy=energy, pk=self.get_descent_direction(energy),
                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

    def get_descent_direction(self, energy):
        """ Calculates the next descent direction.

        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.

        Returns
        -------
        Field
           The descent direction.
        """
        raise NotImplementedError


class SteepestDescent(DescentMinimizer):
    """ Implementation of the steepest descent minimization scheme.

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

    def get_descent_direction(self, energy):
        return -energy.gradient


class RelaxedNewton(DescentMinimizer):
    """ Calculates the descent direction according to a Newton scheme.

    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:
            line_searcher = LineSearchStrongWolfe(
                preferred_initial_step_size=1.)
        super(RelaxedNewton, self).__init__(controller=controller,
                                            line_searcher=line_searcher)

    def get_descent_direction(self, energy):
        return -energy.metric.inverse_times(energy.gradient)


class L_BFGS(DescentMinimizer):
    def __init__(self, controller, line_searcher=LineSearchStrongWolfe(),
                 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

    def get_descent_direction(self, energy):
        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
204
205


206
class VL_BFGS(DescentMinimizer):
Martin Reinecke's avatar
Martin Reinecke committed
207
208
209
210
211
212
213
214
215
216
217
218
219
220
    """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
221
222
    def __init__(self, controller, line_searcher=LineSearchStrongWolfe(),
                 max_history_length=5):
Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
223
224
        super(VL_BFGS, self).__init__(controller=controller,
                                      line_searcher=line_searcher)
225
226
        self.max_history_length = max_history_length

227
    def __call__(self, energy):
228
        self._information_store = None
229
        return super(VL_BFGS, self).__call__(energy)
230

Martin Reinecke's avatar
stage 1    
Martin Reinecke committed
231
232
233
    def reset(self):
        self._information_store = None

234
    def get_descent_direction(self, energy):
235
236
        x = energy.position
        gradient = energy.gradient
237
238
239
240
        # 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
241
242
            self._information_store = _InformationStore(
                self.max_history_length, x0=x, gradient=gradient)
243
244
245
246

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

247
        descent_direction = delta[0] * b[0]
Martin Reinecke's avatar
Martin Reinecke committed
248
        for i in range(1, len(delta)):
249
            descent_direction = descent_direction + delta[i]*b[i]
250

251
        return descent_direction
theos's avatar
theos committed
252
253


Martin Reinecke's avatar
Martin Reinecke committed
254
class _InformationStore(object):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
255
    """Class for storing a list of past updates.
256

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
257
258
    Parameters
    ----------
Martin Reinecke's avatar
Martin Reinecke committed
259
    max_history_length : int
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
260
261
262
263
264
        Maximum number of stored past updates.
    x0 : Field
        Initial position in variable space.
    gradient : Field
        Gradient at position x0.
265

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
266
267
    Attributes
    ----------
Martin Reinecke's avatar
Martin Reinecke committed
268
    max_history_length : int
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
269
270
        Maximum number of stored past updates.
    s : List
Martin Reinecke's avatar
Martin Reinecke committed
271
        Circular buffer of past position differences, which are Fields.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
272
    y : List
Martin Reinecke's avatar
Martin Reinecke committed
273
        Circular buffer of past gradient differences, which are Fields.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
274
    last_x : Field
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
275
        Latest position in variable space.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
276
    last_gradient : Field
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
277
        Gradient at latest position.
Martin Reinecke's avatar
Martin Reinecke committed
278
    k : int
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
279
        Number of updates that have taken place
Martin Reinecke's avatar
Martin Reinecke committed
280
    ss : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
281
        2D circular buffer of scalar products between different elements of s.
Martin Reinecke's avatar
Martin Reinecke committed
282
    sy : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
283
        2D circular buffer of scalar products between elements of s and y.
Martin Reinecke's avatar
Martin Reinecke committed
284
    yy : numpy.ndarray
Martin Reinecke's avatar
Martin Reinecke committed
285
        2D circular buffer of scalar products between different elements of y.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
286
    """
Philipp Arras's avatar
Philipp Arras committed
287

288
289
    def __init__(self, max_history_length, x0, gradient):
        self.max_history_length = max_history_length
290
291
        self.s = [None]*max_history_length
        self.y = [None]*max_history_length
292
293
        self.last_x = x0
        self.last_gradient = gradient
theos's avatar
theos committed
294
        self.k = 0
295

Martin Reinecke's avatar
Martin Reinecke committed
296
        mmax = max_history_length
Martin Reinecke's avatar
Martin Reinecke committed
297
298
299
        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)
300
301
302

    @property
    def history_length(self):
Martin Reinecke's avatar
Martin Reinecke committed
303
        """Returns the number of currently stored updates."""
304
305
306
307
        return min(self.k, self.max_history_length)

    @property
    def b(self):
308
309
        """Combines s, y and gradient to form the new base vectors b.

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
310
311
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
312
        List
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
313
314
            List of new basis vectors.
        """
315
316
        result = []
        m = self.history_length
Martin Reinecke's avatar
Martin Reinecke committed
317
        mmax = self.max_history_length
318

Martin Reinecke's avatar
Martin Reinecke committed
319
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
320
            result.append(self.s[(self.k-m+i) % mmax])
321

Martin Reinecke's avatar
Martin Reinecke committed
322
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
323
            result.append(self.y[(self.k-m+i) % mmax])
324
325
326
327
328
329
330

        result.append(self.last_gradient)

        return result

    @property
    def b_dot_b(self):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
331
        """Generates the (2m+1) * (2m+1) scalar matrix.
332

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
333
        The i,j-th element of the matrix is a scalar product between the i-th
334
335
        and j-th base vector.

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
336
337
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
338
        numpy.ndarray
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
339
340
            Scalar matrix.
        """
341
        m = self.history_length
Martin Reinecke's avatar
Martin Reinecke committed
342
        mmax = self.max_history_length
343
344
345
        k = self.k
        result = np.empty((2*m+1, 2*m+1), dtype=np.float)

Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
346
        # update the stores
Martin Reinecke's avatar
Martin Reinecke committed
347
        k1 = (k-1) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
348
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
349
            kmi = (k-m+i) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
350
351
352
            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
353
        for j in range(m-1):
Martin Reinecke's avatar
Martin Reinecke committed
354
355
            kmj = (k-m+j) % mmax
            self.sy[k1, kmj] = self.s[k1].vdot(self.y[kmj])
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
356

Martin Reinecke's avatar
Martin Reinecke committed
357
        for i in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
358
            kmi = (k-m+i) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
359
            for j in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
360
                kmj = (k-m+j) % mmax
Martin Reinecke's avatar
Martin Reinecke committed
361
362
363
                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]
364

365
            sgrad_i = self.s[kmi].vdot(self.last_gradient)
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
366
            result[2*m, i] = result[i, 2*m] = sgrad_i
367

Martin Reinecke's avatar
fix    
Martin Reinecke committed
368
            ygrad_i = self.y[kmi].vdot(self.last_gradient)
Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
369
            result[2*m, m+i] = result[m+i, 2*m] = ygrad_i
370

Martin Reinecke's avatar
tweaks    
Martin Reinecke committed
371
        result[2*m, 2*m] = self.last_gradient.norm()
372
        return result
theos's avatar
theos committed
373
374

    @property
375
    def delta(self):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
376
        """Calculates the new scalar coefficients (deltas).
377

Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
378
379
        Returns
        -------
Martin Reinecke's avatar
Martin Reinecke committed
380
        List
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
381
382
            List of the new scalar coefficients (deltas).
        """
383
384
385
386
387
388
389
390
        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
391
392
        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)])
393
394
395
            alpha[j] = delta_b_b/b_dot_b[j, m+j]
            delta[m+j] -= alpha[j]

Martin Reinecke's avatar
Martin Reinecke committed
396
        for i in range(2*m+1):
397
398
            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
399
        for j in range(m):
Martin Reinecke's avatar
Martin Reinecke committed
400
            delta_b_b = sum([delta[l]*b_dot_b[m+j, l] for l in range(2*m+1)])
401
402
403
404
405
            beta = delta_b_b/b_dot_b[j, m+j]
            delta[j] += (alpha[j] - beta)

        return delta

theos's avatar
theos committed
406
    def add_new_point(self, x, gradient):
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
407
408
        """Updates the s list and y list.

Martin Reinecke's avatar
Martin Reinecke committed
409
410
        Calculates the new position and gradient differences and enters them
        into the respective list.
Matevz, Sraml (sraml)'s avatar
Matevz, Sraml (sraml) committed
411
        """
Martin Reinecke's avatar
Martin Reinecke committed
412
413
414
        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
415

416
417
        self.last_x = x
        self.last_gradient = gradient
theos's avatar
theos committed
418

Martin Reinecke's avatar
fixes    
Martin Reinecke committed
419
        self.k += 1