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
|
/**
* @file sgd_test.cpp
* @author Ryan Curtin
* @author Marcus Edel
* @author Conrad Sanderson
*
* ensmallen is free software; you may redistribute it and/or modify it under
* the terms of the 3-clause BSD license. You should have received a copy of
* the 3-clause BSD license along with ensmallen. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#if defined(ENS_USE_COOT)
#include <armadillo>
#include <bandicoot>
#endif
#include <ensmallen.hpp>
#include "catch.hpp"
#include "test_function_tools.hpp"
#include "test_types.hpp"
using namespace arma;
using namespace ens;
using namespace ens::test;
template<typename MatType>
void SGDGeneralizedRosenbrockTest(const size_t variants = 50)
{
typedef typename MatType::elem_type ElemType;
const double objTol = 100 * Tolerances<MatType>::Obj;
const double coordTol = 100 * Tolerances<MatType>::Coord;
// Loop over several variants.
for (size_t i = 10; i < variants; i += 5)
{
// Create the generalized Rosenbrock function.
GeneralizedRosenbrockFunction f(i);
// Allow a few trials.
for (size_t trial = 0; trial < 5; ++trial)
{
StandardSGD s(0.001, 1, 1000000, Tolerances<MatType>::Obj / 100, true);
MatType coordinates = f.GetInitialPoint<MatType>();
float result = s.Optimize(f, coordinates);
if (trial != 4)
{
if (result != Approx(0.0).margin(objTol))
continue;
for (size_t j = 0; j < i; ++j)
{
if (ElemType(coordinates(j)) != Approx(1.0).epsilon(coordTol))
continue;
}
}
REQUIRE(result == Approx(0.0).margin(objTol));
for (size_t j = 0; j < i; ++j)
REQUIRE(ElemType(coordinates(j)) == Approx(1.0).epsilon(coordTol));
break;
}
}
}
template<typename MatType>
void SGDLogisticRegressionTest()
{
typedef typename ForwardType<MatType, size_t>::brow LabelsType;
MatType data, testData;
LabelsType responses, testResponses;
LogisticRegressionTestData(data, testData, responses, testResponses);
MatType data2 = data;
LabelsType responses2 = responses;
LogisticRegressionFunction<MatType> lr(data2, responses2, 0.5);
lr.Shuffle();
StandardSGD sgd(0.32);
MatType coordinates = lr.GetInitialPoint();
sgd.Optimize(lr, coordinates);
// Ensure that the error is close to zero.
const double acc = lr.ComputeAccuracy(data, responses, coordinates);
REQUIRE(acc == Approx(100.0).epsilon(Tolerances<MatType>::LRTrainAcc));
const double testAcc = lr.ComputeAccuracy(testData, testResponses,
coordinates);
REQUIRE(testAcc == Approx(100.0).epsilon(Tolerances<MatType>::LRTestAcc));
}
// We skip low precision for this test because tuning SGD for the Rosenbrock
// function is really tricky.
TEMPLATE_TEST_CASE("SGD_GeneralizedRosenbrockFunction", "[SGD]",
ENS_FULLPREC_CPU_TEST_TYPES)
{
SGDGeneralizedRosenbrockTest<TestType>();
}
TEMPLATE_TEST_CASE("SGD_LogisticRegressionFunction", "[SGD]",
ENS_ALL_TEST_TYPES)
{
SGDLogisticRegressionTest<TestType>();
}
|
