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
|
// $Id: bestAlphaManyTrees.cpp 962 2006-11-07 15:13:34Z privmane $
// version 1.00
// last modified 3 Nov 2002
#include "bestAlphaManyTrees.h"
#include "bestAlpha.h"
#include "numRec.h"
#include "bblEMProportional.h"
#include "bblEMSeperate.h"
#include "logFile.h"
#include <iostream>
using namespace std;
#ifndef VERBOS
#define VERBOS
#endif
void bestAlpha::optimizeAlphaNG_EM_PROP(tree& et,
vector<sequenceContainer>& sc,
vector<stochasticProcess>& sp,
const vector<Vdouble *> * weights,
MDOUBLE & bestAlpha,
MDOUBLE & likelihoodScore,
const int maxIterations,
const MDOUBLE epsilon){
//LOG(5,<<" 1. bestAlpha::findBestAlpha"<<endl);
MDOUBLE oldL = VERYSMALL;
MDOUBLE ax,bx,cx; // left, midle, right limit on alpha
bx=1.5; // the limits are becoming more narrow with time.
ax=0;
cx=5.0;
MDOUBLE tol=0.01f;
MDOUBLE bestA=0;
int i;
const int maxIterationsThisF = 50;
for (i=0; i < maxIterationsThisF; ++i) {
bblEMProportional bblEMprop1(et,sc,sp,weights,maxIterations,epsilon);
MDOUBLE newL = bblEMprop1.getTreeLikelihood();
#ifdef VERBOS
LOG(5,<<"Before optimizing alpha, L = "<<newL<<endl);
#endif
MDOUBLE likeAfterAlphaOpt = -brent(ax,bx,cx, // NEW MINUS. CHECK
C_evalAlphaManyTrees(et,sc,sp,weights),
tol,
&bestA); // THIS FUNCTION CHANGE SP, BUT YET ONE HAVE TO INSERT THE BEST ALPHAS.
for (int z=0; z < sp.size();++z) {
(static_cast<gammaDistribution*>(sp[z].distr()))->setAlpha(bestA);
}
#ifdef VERBOS
LOG(5,<<"After optimizing alpha, L = "<<likeAfterAlphaOpt<<endl);
LOG(5,<<" best A = " << bestA<<endl);
#endif
newL = likeAfterAlphaOpt;
if (newL > oldL+0.01) {
oldL = newL;
}
else {
if (newL > oldL) {
likelihoodScore = newL;
bestAlpha= bestA;
return;
}
else {
likelihoodScore = oldL;
bestAlpha= bestA;
return;
}
}
}
if (i == maxIterationsThisF) errorMsg::reportError(" to many iteration in function optimizeBranchLength");
}
/*
void findBestAlphaManyTrees::findBestAlphaFixedManyTrees(const vector<tree>& et,
vector<positionInfo>& pi,
const VVdouble * weights) {
//LOG(5,<<" 1. bestAlpha::findBestAlpha"<<endl);
MDOUBLE bestA=0;
checkAllocation();
MDOUBLE ax,bx,cx; // left, midle, right limit on alpha
MDOUBLE tol;
ax=0;bx=1.5;cx=2;
tol=0.01f;
_bestL = brent(ax,bx,cx,
C_evalAlphaManyTrees(et,_pi,weights),
tol,
&bestA);
_bestAlpha= bestA;
}
*/
void bestAlpha::optimizeAlphaNG_EM_SEP(
vector<tree>& et,
vector<sequenceContainer>& sc,
vector<stochasticProcess>& sp,
const vector<Vdouble *> * weights,
MDOUBLE & bestAlpha,
MDOUBLE & likelihoodScore,
const int maxIterations,
const MDOUBLE epsilon) {
// SEPERATE ANALYSIS, 1 GAMMA
//LOG(5,<<" 1. bestAlpha::findBestAlpha"<<endl);
MDOUBLE oldL = VERYSMALL;
MDOUBLE newL = VERYSMALL;
MDOUBLE ax,bx,cx; // left, midle, right limit on alpha
bx=1.5; // the limits are becoming more narrow with time.
ax=0;
cx=5.0;
MDOUBLE tol=0.01f;
MDOUBLE bestA=0;
const int maxIterationsThisF = 50;
for (int i=0; i < maxIterationsThisF; ++i) {
newL=0;
LOG(3,<<"starting iteration "<<i<<endl);
bblEMSeperate bblEMsep1(et,
sc,
sp,
weights,
maxIterations,
epsilon);
newL =bblEMsep1.getTreeLikelihood();
#ifdef VERBOS
LOG(5,<<"Before optimizing alpha, L = "<<newL<<endl);
#endif
//MDOUBLE alphaB4optimizing = (static_cast<gammaDistribution*>(sp[0].distr()))->getAlpha();
MDOUBLE likeAfterAlphaOpt = -brent(ax,bx,cx, // NEW MINUS - CHECK!
C_evalAlphaManyTreesSep(et,sc,sp,weights),
tol,
&bestA);
if (likeAfterAlphaOpt>newL) {
for (int i=0; i < sc.size();++i) {
(static_cast<gammaDistribution*>(sp[0].distr()))->setAlpha(bestA);
}
newL = likeAfterAlphaOpt;
}
#ifdef VERBOS
LOG(5,<<"After optimizing alpha, L = "<<newL<<endl);
#endif
if (newL > oldL+0.01) {
oldL = newL;
}
else {
if (newL > oldL) {
likelihoodScore = newL;
bestAlpha= bestA;
return;
}
else {
likelihoodScore = oldL;
bestAlpha= bestA;
return;
}
}
}
errorMsg::reportError(" to many iteration in function optimizeBranchLength");
}
//==================== optimizing n alphas ==============================
void bestAlpha::optimizeAlphaNG_EM_PROP_n_alpha(tree& et,
vector<sequenceContainer>& sc,
vector<stochasticProcess>& sp,
const vector<Vdouble *> * weights,
vector<MDOUBLE> & bestAlphas,
MDOUBLE & likelihoodScore,
const int maxIterations,
const MDOUBLE epsilon){
//LOG(5,<<" 1. bestAlpha::findBestAlpha"<<endl);
MDOUBLE oldL = VERYSMALL;
MDOUBLE newL = VERYSMALL;
MDOUBLE ax,bx,cx; // left, midle, right limit on alpha
bx=1.5; // the limits are becoming more narrow with time.
ax=0;
cx=5.0;
vector<MDOUBLE> bestAs= bestAlphas;
vector<MDOUBLE> newAlphas(sc.size(),0);
int i;
const int maxIterationsThisF = 50;
for (i=0; i < maxIterationsThisF; ++i) {
#ifdef VERBOS
LOG(5,<<" ============================ optimizing bbl (fixed alphas) ================= \n");
#endif
newL=0;
bblEMProportional bblem1(et,sc,sp,weights,maxIterations,epsilon);
MDOUBLE tmpX =bblem1.getTreeLikelihood();
#ifdef VERBOS
LOG(5,<<"likelihood of trees (sum)= "<<tmpX<<endl);
#endif
newL =tmpX;
#ifdef VERBOS
LOG(5,<<"Before optimizing alpha, L = "<<newL<<endl);
LOG(5,<<" ============================ optimizing alphas ================= \n");
#endif
const MDOUBLE upperBoundOnAlpha = 5;
MDOUBLE likeAfterAlphaOpt = 0;
for (int treeNumber =0; treeNumber<sc.size();++treeNumber) {
bestAlphaFixedTree bestAlphaFixedTree1(et,
sc[treeNumber],
sp[treeNumber],
weights?(*weights)[treeNumber]:NULL,
upperBoundOnAlpha,
epsilon);
MDOUBLE tmpX = bestAlphaFixedTree1.getBestL();
#ifdef VERBOS
LOG(5,<<"likelihood of tree "<<treeNumber<<" = "<<tmpX<<endl);
#endif
newAlphas[treeNumber] = bestAlphaFixedTree1.getBestAlpha();
#ifdef VERBOS
LOG(5,<<" best alpha tree number: "<<treeNumber<<" = "<<newAlphas[treeNumber]<<endl);
#endif
likeAfterAlphaOpt +=tmpX;
}
if (likeAfterAlphaOpt>newL) {
for (int z=0; z < sp.size();++z) {
(static_cast<gammaDistribution*>(sp[z].distr()))->setAlpha(newAlphas[z]);
}
newL = likeAfterAlphaOpt;
bestAs = newAlphas;
}
#ifdef VERBOS
LOG(5,<<"After optimizing alpha, L = "<<newL<<endl);
#endif
if (newL > oldL+0.01) {
oldL = newL;
}
else {
if (newL > oldL) {
likelihoodScore = newL;
bestAlphas= bestAs;
return;
}
else {
likelihoodScore = oldL;
bestAlphas= bestAs;
return;
}
}
}
if (i == maxIterationsThisF) {
errorMsg::reportError(" to many iteration in function optimizeBranchLength");
}
}
//// CHECK:
//MDOUBLE check_sum=0;
//for (int k=0; k < sp.size(); ++k) {
// MDOUBLE check = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(et,sc[k],sp[k]);
// LOG(5,<<" CHECK = "<< check<<endl);
// check_sum+=check;
//}
//LOG(5,<<" check-sum = "<<check_sum<<endl);
//// END CHECK
|
