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
|
// -*- Mode: C++; tab-width: 2; -*-
// vi: set ts=2:
//
//
#include <BALL/QSAR/validation.h>
#include <BALL/QSAR/statistics.h>
#include <BALL/QSAR/Model.h>
#include <boost/random/mersenne_twister.hpp>
namespace BALL
{
namespace QSAR
{
Validation::Validation(Model* m)
{
model_ = m;
validation_statistic_ = 0;
yRand_results_.resize(0, 0);
}
Validation::~Validation()
{
}
int Validation::getStat() const
{
return validation_statistic_;
}
void Validation::setTrainingLine(int train_line, int current_line)
{
bool fs = 0; // has feature selection being done?
if (!model_->descriptor_IDs_.empty())
{
fs = 1;
}
std::multiset<unsigned int>::iterator it = model_->descriptor_IDs_.begin();
int t = 0; // index in line of training data
// set each cell of the current line
for (unsigned int i = 0; i < model_->data->descriptor_matrix_.size() && (!fs || it != model_->descriptor_IDs_.end()); i++)
{
// set only those cells that belong to selected descriptors
if ( (fs && *it == i) || !fs )
{
// if (train_line == 0)
// {
// model_->transformations(1, t+1) = model_->data->transformations[i][0];
// model_->transformations(2, t+1) = model_->data->transformations[i][1];
// }
model_->descriptor_matrix_(train_line, t) = model_->data->descriptor_matrix_[i][current_line];
t++;
if (fs)
{
it++;
}
}
}
// set all y-values for current substance
// int a = model_->data->transformations.size() - model_->data->Y_.size();
for (unsigned int i = 0; i < model_->data->Y_.size(); i++)
{
model_->Y_(train_line, i) = model_->data->Y_[i][current_line];
// if (train_line == 0)
// {
// transformations(1, col+i+1) = model_->data->transformations[a+i][0];
// transformations(2, col+i+1) = model_->data->transformations[a+i][1];
// }
}
}
void Validation::setTestLine(int test_line, int current_line, bool back_transform)
{
vector<double> v;
test_substances_[test_line] = v;
// COPY ENTIRE LINE!!, relevant descriptors will be automatically chosen by Model.getSubstanceVector(...) (called by Model.predict(...))
for (unsigned int i = 0; i < model_->data->descriptor_matrix_.size(); i++)
{
test_substances_[test_line].push_back(model_->data->descriptor_matrix_[i][current_line]);
if (back_transform)
{
double stddev = model_->data->descriptor_transformations_[i][1];
test_substances_[test_line][i] = test_substances_[test_line][i]*stddev+model_->data->descriptor_transformations_[i][0];
}
}
if (model_->data->y_transformations_.size() == 0)
{
back_transform = 0;
}
// set all y-values for current substance
for (unsigned int i = 0; i < model_->data->Y_.size(); i++)
{
test_Y_(test_line, i) = model_->data->Y_[i][current_line];
if (back_transform)
{
double stddev = model_->data->y_transformations_[i][1];
test_Y_(test_line, i) = test_Y_(test_line, i)*stddev+model_->data->y_transformations_[i][0];
}
}
}
void Validation::yRand()
{
boost::mt19937 rng(PreciseTime::now().getMicroSeconds());
QSARData* data = const_cast <QSARData*> (model_->data);
for (unsigned int i = 0; i < data->Y_.size(); i++)
{
for (unsigned int j = 0; j < data->Y_[0].size(); j++)
{
int pos = rng() % (data->Y_[0].size()-1); // exchange elements at pos and j
double tmp = data->Y_[i][pos];
data->Y_[i][pos] = data->Y_[i][j];
data->Y_[i][j] = tmp;
}
}
}
const Eigen::MatrixXd& Validation::getYRandResults() const
{
return yRand_results_;
}
}
}
|
