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
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
204
205
206
207
208
209
210
211
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
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
|
/* $Id: Idiot.hpp 2078 2015-01-05 12:39:49Z forrest $ */
// Copyright (C) 2002, International Business Machines
// Corporation and others. All Rights Reserved.
// This code is licensed under the terms of the Eclipse Public License (EPL).
// "Idiot" as the name of this algorithm is copylefted. If you want to change
// the name then it should be something equally stupid (but not "Stupid") or
// even better something witty.
#ifndef Idiot_H
#define Idiot_H
#ifndef OSI_IDIOT
#include "ClpSimplex.hpp"
#define OsiSolverInterface ClpSimplex
#else
#include "OsiSolverInterface.hpp"
typedef int CoinBigIndex;
#endif
class CoinMessageHandler;
class CoinMessages;
/// for use internally
typedef struct {
double infeas;
double objval;
double dropThis;
double weighted;
double sumSquared;
double djAtBeginning;
double djAtEnd;
int iteration;
} IdiotResult;
/** This class implements a very silly algorithm. It has no merit
apart from the fact that it gets an approximate solution to
some classes of problems. Better if vaguely homogeneous.
It works on problems where volume algorithm works and often
gets a better primal solution but it has no dual solution.
It can also be used as a "crash" to get a problem started. This
is probably its most useful function.
It is based on the idea that algorithms with terrible convergence
properties may be okay at first. Throw in some random dubious tricks
and the resulting code may be worth keeping as long as you don't
look at it.
*/
class Idiot {
public:
/**@name Constructors and destructor
Just a pointer to model is kept
*/
//@{
/// Default constructor
Idiot ( );
/// Constructor with model
Idiot ( OsiSolverInterface & model );
/// Copy constructor.
Idiot(const Idiot &);
/// Assignment operator. This copies the data
Idiot & operator=(const Idiot & rhs);
/// Destructor
~Idiot ( );
//@}
/**@name Algorithmic calls
*/
//@{
/// Get an approximate solution with the idiot code
void solve();
/// Lightweight "crash"
void crash(int numberPass, CoinMessageHandler * handler,
const CoinMessages * messages, bool doCrossover = true);
/** Use simplex to get an optimal solution
mode is how many steps the simplex crossover should take to
arrive to an extreme point:
0 - chosen,all ever used, all
1 - chosen, all
2 - all
3 - do not do anything - maybe basis
+ 16 do presolves
*/
void crossOver(int mode);
//@}
/**@name Gets and sets of most useful data
*/
//@{
/** Starting weight - small emphasizes feasibility,
default 1.0e-4 */
inline double getStartingWeight() const {
return mu_;
}
inline void setStartingWeight(double value) {
mu_ = value;
}
/** Weight factor - weight multiplied by this when changes,
default 0.333 */
inline double getWeightFactor() const {
return muFactor_;
}
inline void setWeightFactor(double value) {
muFactor_ = value;
}
/** Feasibility tolerance - problem essentially feasible if
individual infeasibilities less than this.
default 0.1 */
inline double getFeasibilityTolerance() const {
return smallInfeas_;
}
inline void setFeasibilityTolerance(double value) {
smallInfeas_ = value;
}
/** Reasonably feasible. Dubious method concentrates more on
objective when sum of infeasibilities less than this.
Very dubious default value of (Number of rows)/20 */
inline double getReasonablyFeasible() const {
return reasonableInfeas_;
}
inline void setReasonablyFeasible(double value) {
reasonableInfeas_ = value;
}
/** Exit infeasibility - exit if sum of infeasibilities less than this.
Default -1.0 (i.e. switched off) */
inline double getExitInfeasibility() const {
return exitFeasibility_;
}
inline void setExitInfeasibility(double value) {
exitFeasibility_ = value;
}
/** Major iterations. stop after this number.
Default 30. Use 2-5 for "crash" 50-100 for serious crunching */
inline int getMajorIterations() const {
return majorIterations_;
}
inline void setMajorIterations(int value) {
majorIterations_ = value;
}
/** Minor iterations. Do this number of tiny steps before
deciding whether to change weights etc.
Default - dubious sqrt(Number of Rows).
Good numbers 105 to 405 say (5 is dubious method of making sure
idiot is not trying to be clever which it may do every 10 minor
iterations) */
inline int getMinorIterations() const {
return maxIts2_;
}
inline void setMinorIterations(int value) {
maxIts2_ = value;
}
// minor iterations for first time
inline int getMinorIterations0() const {
return maxIts_;
}
inline void setMinorIterations0(int value) {
maxIts_ = value;
}
/** Reduce weight after this many major iterations. It may
get reduced before this but this is a maximum.
Default 3. 3-10 plausible. */
inline int getReduceIterations() const {
return maxBigIts_;
}
inline void setReduceIterations(int value) {
maxBigIts_ = value;
}
/// Amount of information - default of 1 should be okay
inline int getLogLevel() const {
return logLevel_;
}
inline void setLogLevel(int value) {
logLevel_ = value;
}
/// How lightweight - 0 not, 1 yes, 2 very lightweight
inline int getLightweight() const {
return lightWeight_;
}
inline void setLightweight(int value) {
lightWeight_ = value;
}
/// strategy
inline int getStrategy() const {
return strategy_;
}
inline void setStrategy(int value) {
strategy_ = value;
}
/// Fine tuning - okay if feasibility drop this factor
inline double getDropEnoughFeasibility() const {
return dropEnoughFeasibility_;
}
inline void setDropEnoughFeasibility(double value) {
dropEnoughFeasibility_ = value;
}
/// Fine tuning - okay if weighted obj drop this factor
inline double getDropEnoughWeighted() const {
return dropEnoughWeighted_;
}
inline void setDropEnoughWeighted(double value) {
dropEnoughWeighted_ = value;
}
/// Set model
inline void setModel(OsiSolverInterface * model) {
model_ = model;
};
//@}
/// Stuff for internal use
private:
/// Does actual work
// allow public!
public:
void solve2(CoinMessageHandler * handler, const CoinMessages *messages);
private:
IdiotResult IdiSolve(
int nrows, int ncols, double * rowsol , double * colsol,
double * pi, double * djs, const double * origcost ,
double * rowlower,
double * rowupper, const double * lower,
const double * upper, const double * element,
const int * row, const CoinBigIndex * colcc,
const int * length, double * lambda,
int maxIts, double mu, double drop,
double maxmin, double offset,
int strategy, double djTol, double djExit, double djFlag,
CoinThreadRandom * randomNumberGenerator);
int dropping(IdiotResult result,
double tolerance,
double small,
int *nbad);
IdiotResult objval(int nrows, int ncols, double * rowsol , double * colsol,
double * pi, double * djs, const double * cost ,
const double * rowlower,
const double * rowupper, const double * lower,
const double * upper, const double * elemnt,
const int * row, const CoinBigIndex * columnStart,
const int * length, int extraBlock, int * rowExtra,
double * solExtra, double * elemExtra, double * upperExtra,
double * costExtra, double weight);
// Deals with whenUsed and slacks
int cleanIteration(int iteration, int ordinaryStart, int ordinaryEnd,
double * colsol, const double * lower, const double * upper,
const double * rowLower, const double * rowUpper,
const double * cost, const double * element, double fixTolerance, double & objChange,
double & infChange, double & maxInfeasibility);
private:
/// Underlying model
OsiSolverInterface * model_;
double djTolerance_;
double mu_; /* starting mu */
double drop_; /* exit if drop over 5 checks less than this */
double muFactor_; /* reduce mu by this */
double stopMu_; /* exit if mu gets smaller than this */
double smallInfeas_; /* feasibility tolerance */
double reasonableInfeas_; /* use lambdas if feasibility less than this */
double exitDrop_; /* candidate for stopping after a major iteration */
double muAtExit_; /* mu on exit */
double exitFeasibility_; /* exit if infeasibility less than this */
double dropEnoughFeasibility_; /* okay if feasibility drop this factor */
double dropEnoughWeighted_; /* okay if weighted obj drop this factor */
int * whenUsed_; /* array to say what was used */
int maxBigIts_; /* always reduce mu after this */
int maxIts_; /* do this many iterations on first go */
int majorIterations_;
int logLevel_;
int logFreq_;
int checkFrequency_; /* can exit after 5 * this iterations (on drop) */
int lambdaIterations_; /* do at least this many lambda iterations */
int maxIts2_; /* do this many iterations on subsequent goes */
int strategy_; /* 0 - default strategy
1 - do accelerator step but be cautious
2 - do not do accelerator step
4 - drop, exitDrop and djTolerance all relative
8 - keep accelerator step to theta=10.0
32 - Scale
512 - crossover
2048 - keep lambda across mu change
4096 - return best solution (not last found)
8192 - always do a presolve in crossover
16384 - costed slacks found - so whenUsed_ longer
32768 - experimental 1
65536 - experimental 2
131072 - experimental 3
262144 - just values pass etc */
int lightWeight_; // 0 - normal, 1 lightweight
};
#endif
|