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
|
// -*- Mode: C++; tab-width: 2; -*-
// vi: set ts=2:
//
//
#include <BALL/QSAR/oplsModel.h>
#include <BALL/QSAR/statistics.h>
namespace BALL
{
namespace QSAR
{
OPLSModel::OPLSModel(const QSARData& q) : PLSModel(q)
{
type_="OPLS";
no_ortho_components_ = 5;
}
OPLSModel::~OPLSModel()
{
}
void OPLSModel::train()
{
Eigen::MatrixXd X = descriptor_matrix_;
// double d[] = {-1, -1, 1, -1, -1, 1, 1, 1};
// //double d[] = {-2.18, -2.18, 1.84, -0.16, -0.48, 1.52, 0.83, 0.83};
// Matrix X(4, 2);
// X << d;
// Y_.ReSize(4, 1);
// double e[] = {2, 2, 0, -4};
// Y_ << e;
int cols = descriptor_matrix_.cols();
Eigen::VectorXd w;
Eigen::VectorXd t;
Eigen::VectorXd c;
Eigen::VectorXd u(X.rows());
for (unsigned int i = 1; i <= u.rows(); i++)
{
u(i) = Y_(i, 1);
}
Eigen::VectorXd u_old;
//w = X.t()*u / Statistics::scalarProduct(u);
//w = w / Statistics::euclNorm(w);
// determine the number of orthogonal components that are to be created.
// no_ortho_components_ contains the number of components desired by the user,
// but obviously the number of orthogonal components must be <= #features-1
// since we will need at least one non-orthogonal component as well
unsigned int orthogonal_components_to_create = no_ortho_components_;
if (cols-1 < no_ortho_components_) orthogonal_components_to_create = cols-1;
W_ortho_.resize(cols, orthogonal_components_to_create);
T_ortho_.resize(descriptor_matrix_.rows(), orthogonal_components_to_create);
for (unsigned int j = 0; j < orthogonal_components_to_create; j++)
{
for (int i = 0; ; i++)
{
w = X.transpose()*u / Statistics::scalarProduct(u);
w = w / Statistics::euclNorm(w);
t = X*w ;
c = Y_.transpose()*t / Statistics::scalarProduct(t);
u_old = u;
u = Y_*c / Statistics::scalarProduct(c);
if (Statistics::euclDistance(u, u_old)/Statistics::euclNorm(u) > 0.0000001)
{
continue;
}
else // if u has converged
{
break;
}
}
Eigen::VectorXd p = X.transpose()*t / Statistics::scalarProduct(t);
Eigen::VectorXd w_ortho = w*((w.transpose()*p) / (w.transpose()*w));
w_ortho = p-w_ortho;
w_ortho = w_ortho / Statistics::euclNorm(w_ortho);
Eigen::VectorXd t_ortho = X*w_ortho / Statistics::scalarProduct(w_ortho);
Eigen::VectorXd p_ortho = X.transpose()*t_ortho / Statistics::scalarProduct(t_ortho);
X -= t_ortho * p_ortho.transpose();
W_ortho_.col(j+1) = w_ortho;
T_ortho_.col(j+1) = t_ortho;
}
descriptor_matrix_ = X;
PLSModel::train();
}
const Eigen::MatrixXd* OPLSModel::getTOrtho()
{
return &T_ortho_;
}
const Eigen::MatrixXd* OPLSModel::getWOrtho()
{
return &W_ortho_;
}
bool OPLSModel::optimizeParameters(int k, int no_steps)
{
double best_q2 = 0;
int best_no = 1;
int best_o_no = 1;
int cols = data->getNoDescriptors();
if (!descriptor_IDs_.empty())
{
cols = descriptor_IDs_.size();
}
no_ortho_components_ = 0;
// first find best number of components, for PLS analysis
for (int i = 1; i <= no_steps && i <= cols; i++)
{
no_components_ = i;
validation->crossValidation(k);
if (validation->getQ2() > best_q2)
{
best_q2 = validation->getQ2();
best_no = i;
}
else if (validation->getQ2() < 0.75*best_q2)
{
break; // for speed-up
}
}
no_components_ = best_no;
//then try to remove orthogonal variance by use of OPLS components
for (int i = 0; i <= no_steps && i <= cols && i < no_components_; i++)
{
no_ortho_components_ = i;
validation->crossValidation(k);
if (validation->getQ2() > best_q2)
{
best_q2 = validation->getQ2();
best_o_no = i;
}
else if (validation->getQ2() < 0.75*best_q2)
{
break; // for speed-up
}
}
no_ortho_components_ = best_o_no;
//finally, try to reduce the number of PLS components (since orthogonal variance has been removed, less PLS components should be neccessary)
for (int i = no_components_; i >= 1 ; i--)
{
no_components_ = i;
validation->crossValidation(k);
if (validation->getQ2() > best_q2)
{
best_q2 = validation->getQ2();
best_no = i;
}
else if (validation->getQ2() < 0.75*best_q2)
{
break; // for speed-up
}
}
no_components_ = best_no;
validation->setQ2(best_q2);
return 1;
}
int OPLSModel::getNoOrthoComponents()
{
return no_ortho_components_;
}
void OPLSModel::setNoOrthoComponents(int d)
{
no_ortho_components_ = d;
}
void OPLSModel::setParameters(vector<double>& v)
{
if (v.size() != 2)
{
String c = "Wrong number of model parameters! Needed: 2;";
c = c+" given: "+String(v.size());
throw Exception::ModelParameterError(__FILE__, __LINE__, c.c_str());
}
no_ortho_components_ = (int)v[0];
no_components_ = (int)v[1];
}
vector<double> OPLSModel::getParameters() const
{
vector<double> d;
d.push_back(no_ortho_components_);
d.push_back(no_components_);
return d;
}
// RowVector OPLSModel::predict(const vector<double> & substance, bool transform)
// {
// if (training_result_.cols() == 0)
// {
// throw Exception::InconsistentUsage(__FILE__, __LINE__, "Model must be trained before it can predict the activitiy of substances!");
// }
//
// RowVector v = getSubstanceVector(substance, transform);
//
// // ColumnVector t_new_ortho = v*W_ortho_.Column(1) / (W_ortho_.Column(1).t()*W_ortho_.Column(1)).AsScalar();
// // v = v - t_new_ortho*P_ortho.Column(1).t();
//
// RowVector res = v*training_result_;
//
// if (transform && y_transformations_.cols() != 0)
// {
// backTransformPrediction(res);
// }
//
// return res;
// }
}
}
|
