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
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
|
// SPDX-License-Identifier: EPL-2.0 OR GPL-2.0-or-later
// SPDX-FileCopyrightText: Bradley M. Bell <bradbell@seanet.com>
// SPDX-FileContributor: 2003-22 Bradley M. Bell
// ----------------------------------------------------------------------------
/*
{xrst_begin rc_sparsity.cpp}
Preferred Sparsity Patterns: Row and Column Indices: Example and Test
#####################################################################
Purpose
*******
This example show how to use row and column index sparsity patterns
:ref:`sparse_rc-name` to compute sparse Jacobians and Hessians.
This became the preferred way to represent sparsity on
:ref:`2017-02-09<2017@mm-dd@02-09>` .
{xrst_literal
// BEGIN C++
// END C++
}
{xrst_end rc_sparsity.cpp}
*/
// BEGIN C++
# include <cppad/cppad.hpp>
namespace {
using CppAD::sparse_rc;
using CppAD::sparse_rcv;
using CppAD::NearEqual;
//
typedef CPPAD_TESTVECTOR(bool) b_vector;
typedef CPPAD_TESTVECTOR(size_t) s_vector;
typedef CPPAD_TESTVECTOR(double) d_vector;
typedef CPPAD_TESTVECTOR( CppAD::AD<double> ) a_vector;
//
double eps99 = 99.0 * std::numeric_limits<double>::epsilon();
// -----------------------------------------------------------------------
// function f(x) that we are computing sparse results for
// -----------------------------------------------------------------------
a_vector fun(const a_vector& x)
{ size_t n = x.size();
a_vector ret(n + 1);
for(size_t i = 0; i < n; i++)
{ size_t j = (i + 1) % n;
ret[i] = x[i] * x[i] * x[j];
}
ret[n] = 0.0;
return ret;
}
// -----------------------------------------------------------------------
// Jacobian
// -----------------------------------------------------------------------
bool check_jac(
const d_vector& x ,
const sparse_rcv<s_vector, d_vector>& subset )
{ bool ok = true;
size_t n = x.size();
//
ok &= subset.nnz() == 2 * n;
const s_vector& row( subset.row() );
const s_vector& col( subset.col() );
const d_vector& val( subset.val() );
s_vector row_major = subset.row_major();
for(size_t i = 0; i < n; i++)
{ size_t j = (i + 1) % n;
size_t k = 2 * i;
//
ok &= row[ row_major[k] ] == i;
ok &= row[ row_major[k+1] ] == i;
//
size_t ck = col[ row_major[k] ];
size_t ckp = col[ row_major[k+1] ];
double vk = val[ row_major[k] ];
double vkp = val[ row_major[k+1] ];
//
// put diagonal element first
if( j < i )
{ std::swap(ck, ckp);
std::swap(vk, vkp);
}
// diagonal element
ok &= ck == i;
ok &= NearEqual( vk, 2.0 * x[i] * x[j], eps99, eps99 );
// off diagonal element
ok &= ckp == j;
ok &= NearEqual( vkp, x[i] * x[i], eps99, eps99 );
}
return ok;
}
// Use forward mode for Jacobian and sparsity pattern
bool forward_jac(CppAD::ADFun<double>& f)
{ bool ok = true;
size_t n = f.Domain();
//
// sparsity pattern for identity matrix
sparse_rc<s_vector> pattern_in(n, n, n);
for(size_t k = 0; k < n; k++)
pattern_in.set(k, k, k);
//
// sparsity pattern for Jacobian
bool transpose = false;
bool dependency = false;
bool internal_bool = false;
sparse_rc<s_vector> pattern_out;
f.for_jac_sparsity(
pattern_in, transpose, dependency, internal_bool, pattern_out
);
//
// compute entire Jacobian
size_t group_max = 1;
std::string coloring = "cppad";
sparse_rcv<s_vector, d_vector> subset( pattern_out );
CppAD::sparse_jac_work work;
d_vector x(n);
for(size_t j = 0; j < n; j++)
x[j] = double(j + 2);
size_t n_sweep = f.sparse_jac_for(
group_max, x, subset, pattern_out, coloring, work
);
//
// check Jacobian
ok &= check_jac(x, subset);
ok &= n_sweep == 2;
//
return ok;
}
// Use reverse mode for Jacobian and sparsity pattern
bool reverse_jac(CppAD::ADFun<double>& f)
{ bool ok = true;
size_t n = f.Domain();
size_t m = f.Range();
//
// sparsity pattern for identity matrix
sparse_rc<s_vector> pattern_in(m, m, m);
for(size_t k = 0; k < m; k++)
pattern_in.set(k, k, k);
//
// sparsity pattern for Jacobian
bool transpose = false;
bool dependency = false;
bool internal_bool = false;
sparse_rc<s_vector> pattern_out;
f.rev_jac_sparsity(
pattern_in, transpose, dependency, internal_bool, pattern_out
);
//
// compute entire Jacobian
std::string coloring = "cppad";
sparse_rcv<s_vector, d_vector> subset( pattern_out );
CppAD::sparse_jac_work work;
d_vector x(n);
for(size_t j = 0; j < n; j++)
x[j] = double(j + 2);
size_t n_sweep = f.sparse_jac_rev(
x, subset, pattern_out, coloring, work
);
//
// check Jacobian
ok &= check_jac(x, subset);
ok &= n_sweep == 2;
//
return ok;
}
// ------------------------------------------------------------------------
// Hessian
// ------------------------------------------------------------------------
bool check_hes(
size_t i ,
const d_vector& x ,
const sparse_rcv<s_vector, d_vector>& subset )
{ bool ok = true;
size_t n = x.size();
size_t j = (i + 1) % n;
//
ok &= subset.nnz() == 3;
const s_vector& row( subset.row() );
const s_vector& col( subset.col() );
const d_vector& val( subset.val() );
s_vector row_major = subset.row_major();
//
double v0 = val[ row_major[0] ];
double v1 = val[ row_major[1] ];
double v2 = val[ row_major[2] ];
if( j < i )
{ ok &= row[ row_major[0] ] == j;
ok &= col[ row_major[0] ] == i;
ok &= NearEqual( v0, 2.0 * x[i], eps99, eps99 );
//
ok &= row[ row_major[1] ] == i;
ok &= col[ row_major[1] ] == j;
ok &= NearEqual( v1, 2.0 * x[i], eps99, eps99 );
//
ok &= row[ row_major[2] ] == i;
ok &= col[ row_major[2] ] == i;
ok &= NearEqual( v2, 2.0 * x[j], eps99, eps99 );
}
else
{ ok &= row[ row_major[0] ] == i;
ok &= col[ row_major[0] ] == i;
ok &= NearEqual( v0, 2.0 * x[j], eps99, eps99 );
//
ok &= row[ row_major[1] ] == i;
ok &= col[ row_major[1] ] == j;
ok &= NearEqual( v1, 2.0 * x[i], eps99, eps99 );
//
ok &= row[ row_major[2] ] == j;
ok &= col[ row_major[2] ] == i;
ok &= NearEqual( v2, 2.0 * x[i], eps99, eps99 );
}
return ok;
}
// Use forward mode for Hessian and sparsity pattern
bool forward_hes(CppAD::ADFun<double>& f)
{ bool ok = true;
size_t n = f.Domain();
size_t m = f.Range();
//
b_vector select_domain(n);
for(size_t j = 0; j < n; j++)
select_domain[j] = true;
sparse_rc<s_vector> pattern_out;
//
for(size_t i = 0; i < m; i++)
{ // select i-th component of range
b_vector select_range(m);
d_vector w(m);
for(size_t k = 0; k < m; k++)
{ select_range[k] = k == i;
w[k] = 0.0;
if( k == i )
w[k] = 1.0;
}
//
bool internal_bool = false;
f.for_hes_sparsity(
select_domain, select_range, internal_bool, pattern_out
);
//
// compute Hessian for i-th component function
std::string coloring = "cppad.symmetric";
sparse_rcv<s_vector, d_vector> subset( pattern_out );
CppAD::sparse_hes_work work;
d_vector x(n);
for(size_t j = 0; j < n; j++)
x[j] = double(j + 2);
size_t n_sweep = f.sparse_hes(
x, w, subset, pattern_out, coloring, work
);
//
// check Hessian
if( i == n )
ok &= subset.nnz() == 0;
else
{ ok &= check_hes(i, x, subset);
ok &= n_sweep == 1;
}
}
return ok;
}
// Use reverse mode for Hessian and sparsity pattern
bool reverse_hes(CppAD::ADFun<double>& f)
{ bool ok = true;
size_t n = f.Domain();
size_t m = f.Range();
//
// n by n identity matrix
sparse_rc<s_vector> pattern_in(n, n, n);
for(size_t j = 0; j < n; j++)
pattern_in.set(j, j, j);
//
bool transpose = false;
bool dependency = false;
bool internal_bool = true;
sparse_rc<s_vector> pattern_out;
//
f.for_jac_sparsity(
pattern_in, transpose, dependency, internal_bool, pattern_out
);
//
for(size_t i = 0; i < m; i++)
{ // select i-th component of range
b_vector select_range(m);
d_vector w(m);
for(size_t k = 0; k < m; k++)
{ select_range[k] = k == i;
w[k] = 0.0;
if( k == i )
w[k] = 1.0;
}
//
f.rev_hes_sparsity(
select_range, transpose, internal_bool, pattern_out
);
//
// compute Hessian for i-th component function
std::string coloring = "cppad.symmetric";
sparse_rcv<s_vector, d_vector> subset( pattern_out );
CppAD::sparse_hes_work work;
d_vector x(n);
for(size_t j = 0; j < n; j++)
x[j] = double(j + 2);
size_t n_sweep = f.sparse_hes(
x, w, subset, pattern_out, coloring, work
);
//
// check Hessian
if( i == n )
ok &= subset.nnz() == 0;
else
{ ok &= check_hes(i, x, subset);
ok &= n_sweep == 1;
}
}
return ok;
}
}
// driver for all of the cases above
bool rc_sparsity(void)
{ bool ok = true;
//
// record the funcion
size_t n = 20;
size_t m = n + 1;
a_vector x(n), y(m);
for(size_t j = 0; j < n; j++)
x[j] = CppAD::AD<double>(j+1);
CppAD::Independent(x);
y = fun(x);
CppAD::ADFun<double> f(x, y);
//
// run the example / tests
ok &= forward_jac(f);
ok &= reverse_jac(f);
ok &= forward_hes(f);
ok &= reverse_hes(f);
//
return ok;
}
// END C++
|
