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
|
// learn.cc
// Ƚ̴ؿˤʹ٤γؽ
#include "osl/progress/ptypeoSquare.h"
#include "osl/progress/ptypeRank.h"
#include "osl/stat/sparseRegressionMultiplier.h"
#include "osl/stat/diagonalPreconditioner.h"
#include "osl/stat/iterativeLinearSolver.h"
#include "osl/stat/twoDimensionalStatistics.h"
#include "osl/record/kisen.h"
#include "osl/numEffectState.h"
#include "osl/effectUtil.h"
#include <deque>
#include <valarray>
#include <cstdlib>
#include <cstdio>
#include <iostream>
#include <unistd.h>
void usage(const char *prog)
{
using namespace std;
cerr << "Usage: " << prog << " [-N#games] [-g]"
<< " [-o weights filename] [-s] [-t tmp weights filename] \n"
<< " -s consider moves.size()\n"
<< " -g gekisashi simulation\n"
<< endl;
exit(1);
}
using namespace osl;
using namespace osl::progress;
using namespace osl::stat;
typedef NumEffectState state_t;
typedef std::valarray<double> valarray_t;
template <class ProgressType>
void processRecord(osl::vector<Move> const& moves);
template <class ProgressType>
void learn(const char *output_filename, const char *tmp_filename);
bool experimental_size_consideration = false;
int main(int argc, char **argv)
{
nice(20);
const char *program_name = argv[0];
bool error_flag = false;
extern char *optarg;
extern int optind;
char c;
size_t maxGames = 0;
const char *output_filename = "weights.txt";
const char *tmp_filename = "tmp_weights.txt";
bool use_ptype_rank = false;
// int infinite_loop = true;
while ((c = getopt(argc, argv, "gN:o:st:vh")) != EOF)
{
switch(c)
{
case 'g': use_ptype_rank = true;
break;
case 'N': maxGames = atoi(optarg);
break;
case 's': experimental_size_consideration = true;
break;
case 't': tmp_filename = optarg;
break;
case 'o': output_filename = optarg;
break;
default: error_flag = true;
}
}
argc -= optind;
argv += optind;
if (error_flag)
usage(program_name);
KisenFile kisenFile("../../data/kisen/01.kif");
if (! maxGames)
maxGames = kisenFile.size();
for (size_t i=0; i<maxGames; ++i)
{
const vector<Move> moves=kisenFile.getMoves(i);
if (use_ptype_rank)
processRecord<PtypeRank>(moves);
else
processRecord<PtypeOSquare>(moves);
}
if (use_ptype_rank)
learn<PtypeRank>(output_filename, tmp_filename);
else
learn<PtypeOSquare>(output_filename, tmp_filename);
return 0;
}
typedef DiscriminationInstance<4> Instance;
static std::deque<Instance> instances;
static std::deque<unsigned short> sizes;
template <class ProgressType>
void processRecord(osl::vector<Move> const& moves)
{
NumEffectState state((SimpleState(HIRATE)));
for (size_t i=0; i<moves.size (); ++i)
{
if (EffectUtil::isKingInCheck(alt(state.turn()), state))
{
// ʬμ֤β => ľμ꤬ˡ
std::cerr << "e"; // state;
break;
}
Instance e;
ProgressType::diff(moves[i], e);
// std::cerr << e << "\n";
instances.push_back(e);
sizes.push_back(moves.size());
state.doMove(moves[i]);
}
}
template <class ProgressType>
struct InstanceMultiplier : public SparseRegressionMultiplier
{
mutable size_t cur;
InstanceMultiplier()
: SparseRegressionMultiplier(ProgressType::maxIndex()), cur(0)
{
}
void printCurrentDate() const
{
struct tm *cur;
time_t ct;
time(&ct);
cur=localtime(&ct);
std::cerr << cur->tm_hour << ":" << cur->tm_min << ":" << cur->tm_sec
<< "\n";
}
void newIteration() const
{
printCurrentDate();
}
bool getVectorX(size_t& num_elements, size_t *non_zero_indices,
double *non_zero_values) const
{
num_elements = instances[cur].size;
for (size_t i=0; i<num_elements; ++i)
{
non_zero_indices[i] = instances[cur].indices[i];
assert(non_zero_indices[i] < dim());
const double value = experimental_size_consideration
? 100.0/sizes[cur] : 1.0;
non_zero_values[i] = (instances[cur].values[i] ? value : -value);
}
cur = (cur+1) % instances.size();
return cur;
}
};
template <class ProgressType>
void learn(const char *output_filename, const char *tmp_filename)
{
std::cerr << instances.size() << "\n";
const size_t dim = ProgressType::maxIndex();
valarray_t result(0.0, dim);
valarray_t b(dim);
valarray_t diag_inv(dim);
InstanceMultiplier<ProgressType> prodA;
std::cerr << "computing x^t y\n";
DoubleConstReader y(1.0);
prodA.computeXtY(y, &b[0], &diag_inv[0]);
DiagonalPreconditioner preconditioner(dim);
preconditioner.setInverseDiagonals(&diag_inv[0]);
std::cerr << "preconditioner\n";
IterativeLinearSolver solver(prodA, &preconditioner, 20, 0.01);
std::cerr << "solver started ";
int err = 0;
int iter;
double tol;
#if 0
std::cerr << "using bicgstab\n";
err = solver.solve_by_BiCGSTAB(b, result, &iter, &tol);
#else
std::cerr << "using cg\n";
err = solver.solve_by_CG(b, result, &iter, &tol);
#endif
std::cerr << "tolerance achieved " << tol << "\n";
std::ofstream os(output_filename);
for (size_t i=0; i<dim; ++i)
os << result[i] << "\n";
}
/* ------------------------------------------------------------------------- */
// ;;; Local Variables:
// ;;; mode:c++
// ;;; c-basic-offset:2
// ;;; End:
|
