aboutsummaryrefslogtreecommitdiffstats
path: root/src/vppinfra/anneal.c
blob: 35d1094648202bb486bf5b8fab08574f6a41021e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
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
/*
  Copyright (c) 2011 Cisco and/or its affiliates.

  * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
*/

#include <vppinfra/anneal.h>

/*
 * Optimize an objective function by simulated annealing
 *
 * Here are a couple of short, easily-understood
 * descriptions of simulated annealing:
 *
 * http://www.cs.sandia.gov/opt/survey/sa.html
 * Numerical Recipes in C, 2nd ed., 444ff
 *
 * The description in the Wikipedia is not helpful.
 *
 * The algorithm tries to produce a decent answer to combinatorially
 * explosive optimization problems by analogy to slow cooling
 * of hot metal, aka annealing.
 *
 * There are (at least) three problem-dependent annealing parameters
 * to consider:
 *
 * t0, the initial "temperature. Should be set so that the probability
 * of accepting a transition to a higher cost configuration is
 * initially about 0.8.
 *
 * ntemps, the number of temperatures to use. Each successive temperature
 * is some fraction of the previous temperature.
 *
 * nmoves_per_temp, the number of configurations to try at each temperature
 *
 * It is a black art to set ntemps, nmoves_per_temp, and the rate
 * at which the temperature drops. Go too fast with too few iterations,
 * and the computation falls into a local minimum instead of the
 * (desired) global minimum.
 */

void
clib_anneal (clib_anneal_param_t * p)
{
  f64 t;
  f64 cost, prev_cost, delta_cost, initial_cost, best_cost;
  f64 random_accept, delta_cost_over_t;
  f64 total_increase = 0.0, average_increase;
  u32 i, j;
  u32 number_of_increases = 0;
  u32 accepted_this_temperature;
  u32 best_saves_this_temperature;
  int accept;

  t = p->initial_temperature;
  best_cost = initial_cost = prev_cost = p->anneal_metric (p->opaque);
  p->anneal_save_best_configuration (p->opaque);

  if (p->flags & CLIB_ANNEAL_VERBOSE)
    fformat (stdout, "Initial cost %.2f\n", initial_cost);

  for (i = 0; i < p->number_of_temperatures; i++)
    {
      accepted_this_temperature = 0;
      best_saves_this_temperature = 0;

      p->anneal_restore_best_configuration (p->opaque);
      cost = best_cost;

      for (j = 0; j < p->number_of_configurations_per_temperature; j++)
	{
	  p->anneal_new_configuration (p->opaque);
	  cost = p->anneal_metric (p->opaque);

	  delta_cost = cost - prev_cost;

	  /* cost function looks better, accept this move */
	  if (p->flags & CLIB_ANNEAL_MINIMIZE)
	    accept = delta_cost < 0.0;
	  else
	    accept = delta_cost > 0.0;

	  if (accept)
	    {
	      if (p->flags & CLIB_ANNEAL_MINIMIZE)
		if (cost < best_cost)
		  {
		    if (p->flags & CLIB_ANNEAL_VERBOSE)
		      fformat (stdout, "New best cost %.2f\n", cost);
		    best_cost = cost;
		    p->anneal_save_best_configuration (p->opaque);
		    best_saves_this_temperature++;
		  }

	      accepted_this_temperature++;
	      prev_cost = cost;
	      continue;
	    }

	  /* cost function worse, keep stats to suggest t0 */
	  total_increase += (p->flags & CLIB_ANNEAL_MINIMIZE) ?
	    delta_cost : -delta_cost;

	  number_of_increases++;

	  /*
	   * Accept a higher cost with Pr { e^(-(delta_cost / T)) },
	   * equivalent to rnd[0,1] < e^(-(delta_cost / T))
	   *
	   * AKA, the Boltzmann factor.
	   */
	  random_accept = random_f64 (&p->random_seed);

	  delta_cost_over_t = delta_cost / t;

	  if (random_accept < exp (-delta_cost_over_t))
	    {
	      accepted_this_temperature++;
	      prev_cost = cost;
	      continue;
	    }
	  p->anneal_restore_previous_configuration (p->opaque);
	}

      if (p->flags & CLIB_ANNEAL_VERBOSE)
	{
	  fformat (stdout, "Temp %.2f, cost %.2f, accepted %d, bests %d\n", t,
		   prev_cost, accepted_this_temperature,
		   best_saves_this_temperature);
	  fformat (stdout, "Improvement %.2f\n", initial_cost - prev_cost);
	  fformat (stdout, "-------------\n");
	}

      t = t * p->temperature_step;
    }

  /*
   * Empirically, one wants the probability of accepting a move
   * at the initial temperature to be about 0.8.
   */
  average_increase = total_increase / (f64) number_of_increases;
  p->suggested_initial_temperature = average_increase / 0.22;	/* 0.22 = -ln (0.8) */

  p->final_temperature = t;
  p->final_metric = p->anneal_metric (p->opaque);

  if (p->flags & CLIB_ANNEAL_VERBOSE)
    {
      fformat (stdout, "Average cost increase from a bad move: %.2f\n",
	       average_increase);
      fformat (stdout, "Suggested t0 = %.2f\n",
	       p->suggested_initial_temperature);
    }
}

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */