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
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
|
/*
Copyright (C) 2011 Tal Pupko TalP@tauex.tau.ac.il.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "gainLossOptimizer.h"
#include "bblEMfixRoot.h"
#include "bblEM.h"
#include "bblLS.h"
/********************************************************************************************
gainLossOptimizer
*********************************************************************************************/
gainLossOptimizer::gainLossOptimizer(tree& tr, stochasticProcess* sp, const sequenceContainer &sc,
const MDOUBLE epsilonOptimization, const int numIterations,
const MDOUBLE epsilonOptimizationModel, const int numIterationsModel,
const MDOUBLE epsilonOptimizationBBL, const int numIterationsBBL,
Vdouble* weights,
unObservableData* unObservableData_p, bool performOptimizationsBBL, bool isbblLSWhenbblEMdontImprove):
_tr(tr),_sp(sp),_sc(sc),
_epsilonOptimization(epsilonOptimization),_maxNumOfIterations(numIterations),
_epsilonOptimizationModel(epsilonOptimizationModel),_maxNumOfIterationsModel(numIterationsModel),
_epsilonOptimizationBBL(epsilonOptimizationBBL),_maxNumOfIterationsBBL(numIterationsBBL),
_weightsUniqPatterns(weights),
_unObservableData_p(unObservableData_p),_performOptimizationsBBL(performOptimizationsBBL),
_isbblLSWhenbblEMdontImprove(isbblLSWhenbblEMdontImprove)
{
//gainLossOptions::distributionType rateDistributionType = getRateDistributionType(sp->distr());
//_weights = gainLossOptions::_weights; // since - no weights are used over positions
_isReversible = !dynamic_cast<gainLossModelNonReversible*>(_sp->getPijAccelerator()->getReplacementModel());
_isSkipBblEM = false; // will change to T if like is not improved by BBL-EM
_freq.resize(_sc.alphabetSize());
optimizations();
}
/********************************************************************************************
*********************************************************************************************/
gainLossOptimizer::gainLossOptimizer(tree& tr, vector<vector<stochasticProcess*> >& spVVec, distribution * gainDist, distribution * lossDist,
const sequenceContainer &sc,
const MDOUBLE epsilonOptimization, const int numIterations,
const MDOUBLE epsilonOptimizationModel, const int numIterationsModel,
const MDOUBLE epsilonOptimizationBBL, const int numIterationsBBL,
Vdouble* weights,
unObservableData* unObservableData_p, bool performOptimizationsBBL, bool isbblLSWhenbblEMdontImprove):
_tr(tr),_spVVec(spVVec),_gainDist(gainDist),_lossDist(lossDist),
_sc(sc),//_spSimple(spSimple), // ignore sent model, make new one
_epsilonOptimization(epsilonOptimization),_maxNumOfIterations(numIterations),
_epsilonOptimizationModel(epsilonOptimizationModel),_maxNumOfIterationsModel(numIterationsModel),
_epsilonOptimizationBBL(epsilonOptimizationBBL),_maxNumOfIterationsBBL(numIterationsBBL),
_weightsUniqPatterns(weights),
_unObservableData_p(unObservableData_p),_performOptimizationsBBL(performOptimizationsBBL),
_isbblLSWhenbblEMdontImprove(isbblLSWhenbblEMdontImprove)
{
//_sp = _spVVec[0][0]; //used for reference (Alpha and such)
//_weights = gainLossOptions::_weights; // since - no weights are used over positions
_spSimple = NULL;
_isSkipBblEM = false; // will change to T if like is not improved by BBL-EM
_freq.resize(_sc.alphabetSize());
_bestGainBeta = 1;
_bestLossBeta = 1;
optimizationsSPvv();
}
/********************************************************************************************
optimizations
*********************************************************************************************/
void gainLossOptimizer::optimizations(){
time_t t1;
time(&t1);
time_t t2;
LOGnOUT(4,<<"-------------------------------"<<endl
<<"Starting optimizations: maxNumIterations="<<_maxNumOfIterations<<endl);
_bestL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
MDOUBLE currBestL=VERYSMALL;
MDOUBLE previousL;
bool noLikeImprovmentAtBBL = false; // if BBL did not produce a new tree, end. (no point in another iteration model+BBL)
bool isSkipParamsOptimization = gainLossOptions::_isSkipFirstParamsOptimization;
LOGnOUT(3,<<endl<<"#########"<<" optimization starting epsilonCycle="<<_epsilonOptimization<<" maxNumIterations="<<_maxNumOfIterations<<endl);
LOGnOUT(3,<<"start optimization with L= "<<_bestL<<endl);
int iter;
for (iter=1;iter<=_maxNumOfIterations;iter++){
LOGnOUT(4,<<endl<<"------"<<" Model+BBL iter="<<iter<<endl);
previousL = _bestL; // breaking out of loop when no (>epsilon) improvement is made by comparing to previousL
// model optimization
if(!isSkipParamsOptimization){
currBestL = optimizeParameters();}
else{
LOGnOUT(4,<<"Optimize Params - Skipped"<<endl);}
if (currBestL>_bestL) {
_bestL = currBestL;
}
else if(!isSkipParamsOptimization && currBestL<_bestL){
LOGnOUT(4,<<" !!! Warning !!!: after model optimization likelihood went down"<< currBestL<<" "<<_bestL<<endl);
}
//_bestL = max(currBestL,_bestL);
isSkipParamsOptimization = false; // only first iteration skipped
// BBL optimization
if (gainLossOptions::_performOptimizationsBBL && _performOptimizationsBBL) // we use the && 2 enable optimizationsManyStarts not to perform BBL
{
//LOGnOUT(4,<<"Start BBL... with epsilonOptimizationBBL= "<<_epsilonOptimizationBBL<<endl);
if(gainLossOptions::_performOptimizationsBBLOnlyOnce) // Next iteration - no BBL
_performOptimizationsBBL = false;
currBestL = optimizeBranchLengths(iter);
if (currBestL>_bestL) {
_bestL = currBestL;
}
else{
noLikeImprovmentAtBBL = true;
LOGnOUT(4,<<" !!! Warning !!!: after BBL likelihood did not improve"<< currBestL<<" "<<_bestL<<endl);
}
//_bestL = max(currBestL,_bestL);
string treeINodes = gainLossOptions::_outDir + "//" + "TheTree.INodes.iter"+ int2string(iter) + ".ph";
printTree (_tr, treeINodes);
}
// ROOT optimization
if (gainLossOptions::_performOptimizationsROOT)
{
currBestL = optimizeRoot();
if (currBestL>_bestL) {
_bestL = currBestL;
}
else{
LOGnOUT(4,<<" !!! Warning !!!: after Root likelihood did not improve"<< currBestL<<" "<<_bestL<<endl);
}
//_bestL = max(currBestL,_bestL);
}
if ( (_bestL-previousL) < max(_epsilonOptimization, abs(_bestL/10000)) || noLikeImprovmentAtBBL) // stop Opt for less than epsilon likelihood point
{
LOGnOUT(3,<<" OverAll optimization converged. Iter= "<<iter<<" Likelihood="<<_bestL<<endl);
break;
}
if(gainLossOptions::_simulatedAnnealing){
_epsilonOptimization = max(_epsilonOptimization*gainLossOptions::_simulatedAnnealingCoolingFactor,0.3*gainLossOptions::_simulatedAnnealingMinEpsilonFactor); // simulated annealing
_epsilonOptimizationModel = max(_epsilonOptimizationModel*gainLossOptions::_simulatedAnnealingCoolingFactor,0.1*gainLossOptions::_simulatedAnnealingMinEpsilonFactor); // simulated annealing
_epsilonOptimizationBBL = max(_epsilonOptimizationBBL*gainLossOptions::_simulatedAnnealingCoolingFactor,0.2*gainLossOptions::_simulatedAnnealingMinEpsilonFactor); // simulated annealing
}
}
if (iter>_maxNumOfIterations)
LOGnOUT(4,<<" Too many="<<iter-1<<" iterations in Model+BBL. Last optimized parameters are used."<<endl);
time(&t2);
LOGnOUT(4,<<"Optimization RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
}
/********************************************************************************************
optimizationsSPvv
*********************************************************************************************/
void gainLossOptimizer::optimizationsSPvv(){
time_t t1;
time(&t1);
time_t t2;
LOGnOUT(4,<<"-------------------------------"<<endl
<<"Starting optimizations: maxNumIterations="<<_maxNumOfIterations<<endl);
_bestL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
MDOUBLE currBestL=VERYSMALL;
MDOUBLE previousL;
bool noLikeImprovmentAtBBL = false; // if BBL did not produce a new tree, end. (no point in another iteration model+BBL)
bool isSkipParamsOptimization = gainLossOptions::_isSkipFirstParamsOptimization;
int iter;
LOGnOUT(3,<<endl<<"#########"<<" optimization starting epsilonCycle="<<_epsilonOptimization<<" maxNumIterations="<<_maxNumOfIterations<<endl);
LOGnOUT(3,<<"start optimization with L= "<<_bestL<<endl);
for (iter=1;iter<=_maxNumOfIterations;iter++){
LOGnOUT(4,<<endl<<"------"<<" Model+BBL iter="<<iter<<endl);
previousL = _bestL; // breaking out of loop when no (>epsilon) improvement is made by comparing to previousL
// model optimization
if(!isSkipParamsOptimization){
currBestL = optimizeParametersSPvv();}
else{
LOGnOUT(4,<<"Optimize Params - Skipped"<<endl);
}
if (currBestL>_bestL) {
_bestL = currBestL;
}
else if(!isSkipParamsOptimization && currBestL<_bestL){
LOGnOUT(4,<<" !!! Warning !!!: after model optimization likelihood went down"<< currBestL<<" "<<_bestL<<endl);
}
//_bestL = max(currBestL,_bestL);
isSkipParamsOptimization = false; // only first iteration skipped
// ROOT optimization
if (gainLossOptions::_performOptimizationsROOT)
{
currBestL = optimizeRootSPvv();
if (currBestL>_bestL) {
_bestL = currBestL;
}
else{
LOGnOUT(4,<<" !!! Warning !!!: after Root likelihood did not improve"<< currBestL<<" "<<_bestL<<endl);
}
//_bestL = max(currBestL,_bestL);
}
// BBL optimization
if (gainLossOptions::_performOptimizationsBBL && _performOptimizationsBBL){
//LOGnOUT(4,<<"Start BBL... with epsilonOptimizationBBL= "<<_epsilonOptimizationBBL<<endl);
if(gainLossOptions::_performOptimizationsBBLOnlyOnce) // Next iteration - no BBL
_performOptimizationsBBL = false;
currBestL = optimizeBranchLengthsvv(iter);
if (currBestL>_bestL) {
_bestL = currBestL;
}
else{
noLikeImprovmentAtBBL = true;
LOGnOUT(4,<<" !!! Warning !!!: after BBL likelihood did not improve"<< currBestL<<" "<<_bestL<<endl);
}
//_bestL = max(currBestL,_bestL);
string treeINodes = gainLossOptions::_outDir + "//" + "TheTree.INodes.iter"+ int2string(iter) + ".ph";
printTree (_tr, treeINodes);
}
if ((_bestL-previousL) < max(_epsilonOptimization, abs(_bestL/10000)) || noLikeImprovmentAtBBL ) // stop Opt for less than 2 likelihood point
{
LOGnOUT(3,<<" OverAll optimization converged. Iter= "<<iter<<" Likelihood="<<_bestL<<endl);
break;
}
if(gainLossOptions::_simulatedAnnealing){
_epsilonOptimization = max(_epsilonOptimization*gainLossOptions::_simulatedAnnealingCoolingFactor,0.3*gainLossOptions::_simulatedAnnealingMinEpsilonFactor); // simulated annealing
_epsilonOptimizationModel = max(_epsilonOptimizationModel*gainLossOptions::_simulatedAnnealingCoolingFactor,0.1*gainLossOptions::_simulatedAnnealingMinEpsilonFactor); // simulated annealing
_epsilonOptimizationBBL = max(_epsilonOptimizationBBL*gainLossOptions::_simulatedAnnealingCoolingFactor,0.2*gainLossOptions::_simulatedAnnealingMinEpsilonFactor); // simulated annealing
}
}
if (iter>_maxNumOfIterations)
LOGnOUT(4,<<" Too many="<<iter-1<<" iterations in Model+BBL. Last optimized parameters are used."<<endl);
time(&t2);
LOGnOUT(4,<<"Optimization RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
}
/********************************************************************************************
optimizeParameters
*********************************************************************************************/
MDOUBLE gainLossOptimizer::optimizeParameters(){
//LOGnOUT(4,<<"Starting optimizeParameters with: numOfIterations="<<_maxNumOfIterations<<" and epsilonOptimization="<<_epsilonOptimizationModel<<endl);
time_t t1;
time(&t1);
time_t t2;
optimizeGainLossModel* opt = new optimizeGainLossModel(_tr,*_sp,_sc,
_isReversible,_epsilonOptimizationModel,_maxNumOfIterationsModel,_weightsUniqPatterns,_unObservableData_p);
//optimizeGainLossModel* opt = new optimizeGainLossModel(_tr,*_sp,_sc,
// _isReversible,_epsilonOptimizationModel,_maxNumOfIterationsModel,_plogLforMissingData);
LOGnOUT(4,<<"-------------------------------"<<endl<<"Model optimization over with: "<<endl);
_bestGain=opt->getBestMu1();
LOGnOUT(4,<<"Gain "<<_bestGain<<endl);
//if (!gainLossOptions::_isReversible) {
//MDOUBLE bestM2=opt->getBestMu2();
_bestLoss = static_cast<gainLossModelNonReversible*>(_sp->getPijAccelerator()->getReplacementModel())->getMu2();
LOGnOUT(4,<<"Loss "<<_bestLoss<<endl);
LOGnOUT(4,<<" Gain/Loss ratio= "<< _bestGain/_bestLoss<<endl);
//}
if(isAlphaOptimization((*_sp).distr())){
_bestAlphaRate=opt->getBestAlpha();
LOGnOUT(4,<<"AlphaRateRate "<<_bestAlphaRate<<endl);
}
if(isBetaOptimization((*_sp).distr())){
_bestBetaRate=opt->getBestBeta();
LOGnOUT(4,<<"BetaRate "<<_bestBetaRate<<endl);
LOGnOUT(4,<<" Rate Expectancy = "<< _bestAlphaRate/_bestBetaRate<<endl);
LOGnOUT(4,<<" Rate Standatd Deviation = "<< sqrt(_bestAlphaRate/(_bestAlphaRate*_bestBetaRate))<<endl);
}
if(isInvariantOptimization((*_sp).distr())){
_bestRateProbInvariant =opt->getBestRateProbInvariant();
LOGnOUT(4,<<"ProbInvariantRate "<<_bestRateProbInvariant<<endl);
}
if(dynamic_cast<mixtureDistribution*>((*_sp).distr())){
printMixtureParams();
}
if (isThetaOptimization() && !gainLossOptions::_isRootFreqEQstationary) {
_bestTheta=opt->getBestTheta();
LOGnOUT(4,<<"Theta "<<_bestTheta<<endl);
_freq[1] = _bestTheta;
_freq[0] = 1-_freq[1];
}
else{
_freq[1] = _bestGain/(_bestGain+_bestLoss);
_freq[0] = 1-_freq[1];
}
MDOUBLE bestL = opt->getBestL();
MDOUBLE currentlogL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
if(!DEQUAL(currentlogL,bestL)){ //
LOGnOUT(3,<<"!!! ERROR: different likelihood after optimizeGainLossModel,diff= "<<currentlogL-bestL <<"\n");
}
LOGnOUT(4,<<"updated likelihood (after optimizeParameters)= "<<bestL<<endl);
time(&t2);
LOGnOUT(4,<<"Model optimization RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
if(opt) delete opt;
return bestL;
}
/********************************************************************************************
optimizeParametersSPvv
*********************************************************************************************/
MDOUBLE gainLossOptimizer::optimizeParametersSPvv(){
//LOGnOUT(4,<<"Starting optimizeParametersSPvv with: numOfIterations="<<_maxNumOfIterationsModel<<" and epsilonOptimization="<<_epsilonOptimizationModel<<endl);
time_t t1;
time(&t1);
time_t t2;
optimizeGainLossModelVV* opt = new optimizeGainLossModelVV(_tr,_spVVec,_sc,_gainDist,_lossDist,
gainLossOptions::_isReversible,_epsilonOptimizationModel,_maxNumOfIterationsModel,_weightsUniqPatterns,_unObservableData_p);
LOGnOUT(4,<<"-------------------------------"<<endl<<"Model optimization over with: "<<endl);
_bestGainAlpha=opt->getBestGainAlpha();
LOGnOUT(4,<<"AlphaGain "<<_bestGainAlpha<<endl);
if(isBetaOptimization(_gainDist)){
_bestGainBeta=opt->getBestGainBeta();
LOGnOUT(4,<<"BetaGain "<<_bestGainBeta<<endl);
_gainExp = _bestGainAlpha/_bestGainBeta;
_gainSTD = sqrt(_bestGainAlpha/(_bestGainBeta*_bestGainBeta));
LOGnOUT(4,<<" Gain Expectation = "<< rateExpectation(_gainDist)<<endl);
LOGnOUT(4,<<" Gain Expectancy = "<< _gainExp<<endl);
LOGnOUT(4,<<" Gain Standard Deviation = "<<_gainSTD<<endl);
}
if(isInvariantOptimization(_gainDist)){
_bestGainProbInvariant=opt->getBestGainProbInvariant();
LOGnOUT(4,<<"ProbInvariantGain "<<_bestGainProbInvariant<<endl);
}
if (!gainLossOptions::_isReversible) {
_bestLossAlpha=opt->getBestLossAlpha();
LOGnOUT(4,<<"AlphaLoss "<<_bestLossAlpha<<endl);
if(isBetaOptimization(_lossDist)){
_bestLossBeta=opt->getBestLossBeta();
_lossExp = _bestLossAlpha/_bestLossBeta;
_lossSTD = sqrt(_bestLossAlpha/(_bestLossBeta*_bestLossBeta));
LOGnOUT(4,<<"BetaLoss "<<_bestLossBeta<<endl);
LOGnOUT(4,<<" Loss Expectation = "<< rateExpectation(_lossDist)<<endl);
LOGnOUT(4,<<" Loss Expectancy = "<< _lossExp<<endl);
LOGnOUT(4,<<" Loss Standard Deviation = "<<_lossSTD<<endl);
LOGnOUT(4,<<" Gain/Loss Expectancy ratio= "<< (_bestGainAlpha/_bestGainBeta)/(_bestLossAlpha/_bestLossBeta)<<endl);
LOGnOUT(4,<<" Expectancy(Gain)/Expectancy(Loss) by computation = "<< computeExpOfGainByExpOfLossRatio(_gainDist, _lossDist)<<endl);
}
if(isInvariantOptimization(_lossDist)){
_bestLossProbInvariant=opt->getBestLossProbInvariant();
LOGnOUT(4,<<"ProbInvariantLoss "<<_bestLossProbInvariant<<endl);
}
}
if(isAlphaOptimization((*_spVVec[0][0]).distr())){
MDOUBLE bestAlpha=opt->getBestRateAlpha();
LOGnOUT(4,<<"AlphaRate "<<bestAlpha<<endl);
}
if(isInvariantOptimization((*_spVVec[0][0]).distr())){
MDOUBLE bestRateProbInvariant =opt->getBestRateProbInvariant();
LOGnOUT(4,<<"ProbInvariantRate "<<bestRateProbInvariant<<endl);
}
if (isThetaOptimization() && !gainLossOptions::_isRootFreqEQstationary) {
_bestTheta=opt->getBestTheta();
LOGnOUT(4,<<"Theta "<<_bestTheta<<endl);
_freq[1] = _bestTheta;
_freq[0] = 1-_freq[1];
}
else{
_freq[1] = _gainExp/(_gainExp+_lossExp);
_freq[0] = 1-_freq[1];
}
MDOUBLE bestL = opt->getBestL();
//if(_unObservableData_p) _unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
MDOUBLE currentlogL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
if(!DEQUAL(currentlogL,bestL)){ //DEQUAL(currentlogL,bestL)
LOGnOUT(3,<<"!!! ERROR: different likelihood after optimizeGainLossModel,diff= "<<currentlogL-bestL <<"\n");
}
LOGnOUT(4,<<"updated likelihood (after optimizeParameters)= "<<bestL<<endl);
time(&t2);
LOGnOUT(4,<<"Model optimization RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
if(opt) delete opt;
// update spSimple based on latest est parameters - deleted at the end
if(_spSimple) delete _spSimple;
_spSimple = startStochasticProcessSimpleGamma(_gainExp, _lossExp, _freq);
return bestL;
}
/********************************************************************************************
optimizeRoot
*********************************************************************************************/
MDOUBLE gainLossOptimizer::optimizeRootSPvv()
{
time_t t1;
time(&t1);
time_t t2;
MDOUBLE oldL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
MDOUBLE newL = VERYSMALL;
tree tempTree = _tr;
LOGnOUT(4,<<"*** Starting optimizeRoot="<<oldL<<endl);
treeIterDownTopConst tIt(tempTree);
for (tree::nodeP mynode = tIt.first(); mynode != tIt.end(); mynode = tIt.next()) {
if(mynode->isLeaf())
continue;
tempTree.rootAt(mynode);
if(_unObservableData_p) _unObservableData_p->setLforMissingData(tempTree,_spVVec,_gainDist,_lossDist);
MDOUBLE newL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(tempTree,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
if((newL>oldL+_epsilonOptimizationBBL*gainLossOptions::_epsilonForReRootFactor)){ // only for substantial improvement the tree will be re-rooted
_tr = tempTree;
LOGnOUT(4,<<"tree rooted at "<<_tr.getRoot()->name()<<" with lL="<<newL<<endl);
LOGnOUT(4,<<"sons of root are "<<_tr.getRoot()->getSon(0)->name()<<" , "<<_tr.getRoot()->getSon(1)->name()<<" , "<<_tr.getRoot()->getSon(2)->name()<<endl);
}
else{
if(_unObservableData_p) _unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
}
}
LOGnOUT(4,<<"*** After optimizeRoot="<<max(newL,oldL)<<endl);
time(&t2);
LOGnOUT(4,<<"optimizeRoot RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
return newL;
}
/********************************************************************************************
optimizeRoot
*********************************************************************************************/
MDOUBLE gainLossOptimizer::optimizeRoot()
{
time_t t1;
time(&t1);
time_t t2;
MDOUBLE oldL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
MDOUBLE newL = VERYSMALL;
tree tempTree = _tr;
LOGnOUT(4,<<"*** Starting optimizeRoot= "<<oldL<<endl);
treeIterDownTopConst tIt(tempTree);
for (tree::nodeP mynode = tIt.first(); mynode != tIt.end(); mynode = tIt.next()) {
if(mynode->isLeaf())
continue;
tempTree.rootAt(mynode);
if(_unObservableData_p){
_unObservableData_p->setLforMissingData(tempTree,_sp);
}
MDOUBLE newL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(tempTree,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
if((newL>oldL+_epsilonOptimizationBBL*gainLossOptions::_epsilonForReRootFactor))// only for substantial improvement the tree will be re-rooted
{
_tr = tempTree;
LOGnOUT(4,<<"tree re-rooted at "<<_tr.getRoot()->name()<<" with lL="<<newL<<endl);
LOGnOUT(4,<<"sons of root are "<<_tr.getRoot()->getSon(0)->name()<<" , "<<_tr.getRoot()->getSon(1)->name()<<" , "<<_tr.getRoot()->getSon(2)->name()<<endl);
}
else{
if(_unObservableData_p) _unObservableData_p->setLforMissingData(_tr,_sp); // go back to original tree value
}
}
LOGnOUT(4,<<"*** After optimizeRoot= "<<max(newL,oldL)<<endl);
time(&t2);
LOGnOUT(4,<<"optimizeRoot RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
return newL;
}
/********************************************************************************************
optimizeBranchLengths
*********************************************************************************************/
MDOUBLE gainLossOptimizer::optimizeBranchLengths(const int outerIter)
{
time_t t1;
time(&t1);
time_t t2;
MDOUBLE tollForPairwiseDist=0.01; // the BBL default, epsilon per branch (brent's value)
MDOUBLE bblEMfactor = 10;
int numberOfBranchs = _tr.getNodesNum();
MDOUBLE epsilonOptimizationIterFactor = numberOfBranchs/5; // (is 1.5) for 100 branches (~50 species) the epsilon for the entire iter is 50 times the one for branch
epsilonOptimizationIterFactor = max(5.0,epsilonOptimizationIterFactor);
MDOUBLE epsilonOptimizationBBLIter = _epsilonOptimizationBBL*epsilonOptimizationIterFactor/bblEMfactor; // The next iteration epsilon, multiply per-branch value
MDOUBLE oldL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
MDOUBLE newLnoUnObservableDataCorrection = VERYSMALL;
MDOUBLE newL = VERYSMALL;
tree oldTree = _tr;
bool isFixedRoot = !_isReversible && !gainLossOptions::_isRootFreqEQstationary;
MDOUBLE minLikeImprovmentForNoLS = 5.0;
MDOUBLE minLikeImprovmentForNoSkip = 2.0;
if(gainLossOptions::_isBblLS){
if (gainLossOptions::_isBblEMbeforeLSWithMissSpecifiedModel) {
// start with BBL-EM and additional iteration of Line-Search optimizeBranches
MDOUBLE oldLnoUnObservableDataCorrection = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,NULL);
LOGnOUT(4,<<"*** Prior BBL-EM... followed by addtional iteration of Line-Search "<<"\t"<<oldL<<endl);
if(isFixedRoot){
LOGnOUT(4,<<"*** Start Fix Root BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
bblEMfixRoot bblEM1(_tr, _sc, *_sp, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
,NULL,NULL); // optional &oldLnoUnObservableDataCorrection
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_sp);
newL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
}
else{
LOGnOUT(4,<<"*** Start BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
// Note: likelihood does not improve with iterations compared to the likelihood under correction for UnObs, hence NULL
bblEM bblEM1(_tr, _sc, *_sp, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
,NULL,NULL); // optional &oldLnoUnObservableDataCorrection
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_sp);
newL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
}
bblLS bbl;
MDOUBLE bestL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
newL = bbl.optimizeBranches(_tr,_sp,_sc,_weightsUniqPatterns,_unObservableData_p,outerIter,_epsilonOptimizationBBL, 1 ,bestL);
if(newL<oldL){
_tr = oldTree;
LOGnOUT(4,<<"NOTE: No improvment-> Retain previous tree"<<endl);
}
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
}
// only Line-Search optimizeBranches
else{
bblLS bbl;
MDOUBLE bestL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
newL = bbl.optimizeBranches(_tr,_sp,_sc,_weightsUniqPatterns,_unObservableData_p,outerIter,_epsilonOptimizationBBL,_maxNumOfIterationsBBL,bestL);
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
}
}
else{
if(!_isSkipBblEM){
MDOUBLE oldLnoUnObservableDataCorrection = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,NULL);
if(isFixedRoot){
LOGnOUT(4,<<"*** Start Fix Root BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
bblEMfixRoot bblEM1(_tr, _sc, *_sp, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
,NULL,NULL); // optional &oldLnoUnObservableDataCorrection
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_sp);
newL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
}
else{
LOGnOUT(4,<<"*** Start BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
// Note: likelihood does not improve with iterations compared to the likelihood under correction for UnObs, hence NULL
bblEM bblEM1(_tr, _sc, *_sp, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
,NULL,NULL); // optional &oldLnoUnObservableDataCorrection
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_sp);
newL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
}
}
// if include LS when BBL-EM fails
if((newL-oldL < max(minLikeImprovmentForNoLS, abs(newL/10000)) ) && _isbblLSWhenbblEMdontImprove){ // Do LS if less than 5 likelihood points were gained
LOGnOUT(4,<<" Only "<< newL-oldL<<" improvement with BBL-EM -> Perform BBL-LS one iteration"<<endl);
if(gainLossOptions::_isSkipBblEMWhenbblEMdontImprove && (newL-oldL < minLikeImprovmentForNoSkip)){
LOGnOUT(4,<<"Since no improvement (less than "<<minLikeImprovmentForNoSkip<<"), BBL-EM will be skipped next iteration, go directly to LS "<<endl);
_isSkipBblEM = true; // once BBL-EM is not improving Like, next time - skip
}
_tr = oldTree;
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_sp);
bblLS bbl;
MDOUBLE bestL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
newL = bbl.optimizeBranches(_tr,_sp,_sc,_weightsUniqPatterns,_unObservableData_p,outerIter,_epsilonOptimizationBBL, 1 ,bestL);
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
}
}
LOGnOUT(4,<<"*** After Branch Lengths Opt. returned Likelihood="<<"\t"<<newL<<endl);
if((newL<oldL)){
_tr = oldTree;
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_sp);
newL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
LOGnOUT(4,<<"NOTE: No improvment-> Retain previous tree"<<endl);
}
MDOUBLE postL = likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,*_sp,_weightsUniqPatterns,_unObservableData_p);
if( !DEQUAL(newL,postL) ){
LOGnOUT(3,<<"***ERROR***: Diff returned L, and re-calculated L"<<" "<<newL<<" "<<postL<<" "<<postL-newL<<endl);
}
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
time(&t2);
LOGnOUT(4,<<"Branch Lengths RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
return newL;
}
/********************************************************************************************
optimizeBranchLengthsvv - new
*********************************************************************************************/
MDOUBLE gainLossOptimizer::optimizeBranchLengthsvv(const int outerIter)
{
time_t t1;
time(&t1);
time_t t2;
MDOUBLE tollForPairwiseDist=0.01; // the BBL default, epsilon per branch (brent's value)
MDOUBLE bblEMfactor = 10;
int numberOfBranchs = _tr.getNodesNum();
MDOUBLE epsilonOptimizationIterFactor = numberOfBranchs/5; // (is 1.5) for 100 branches (~50 species) the epsilon for the entire iter is 50 times the one for branch
epsilonOptimizationIterFactor = max(5.0,epsilonOptimizationIterFactor);
MDOUBLE epsilonOptimizationBBLIter = _epsilonOptimizationBBL*epsilonOptimizationIterFactor/bblEMfactor; // The next iteration epsilon, multiply per-branch value
MDOUBLE oldL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
MDOUBLE newLnoUnObservableDataCorrection = VERYSMALL;
MDOUBLE newL = VERYSMALL;
tree oldTree = _tr;
//bool isFixedRoot = !_isReversible && !gainLossOptions::_isRootFreqEQstationary;
MDOUBLE minLikeImprovmentForNoLS = 5.0;
MDOUBLE minLikeImprovmentForNoSkip = 2.0;
if(gainLossOptions::_isBblLS){
if (gainLossOptions::_isBblEMbeforeLSWithMissSpecifiedModel) {
// start with BBL-EM and additional iteration of Line-Search optimizeBranches
MDOUBLE oldLnoUnObservableDataCorrection = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,NULL);
LOGnOUT(4,<<"*** Prior BBL-EM... followed by addtional iteration of Line-Search "<<"\t"<<oldL<<endl);
//if(isFixedRoot){
// LOGnOUT(4,<<"*** spSimple - Fix Root BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
// LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
// bblEMfixRoot bblEM1(_tr, _sc, *_spSimple, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
// ,NULL,&oldLnoUnObservableDataCorrection);
// newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
// if(_unObservableData_p)
// _unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
// newL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
//}
//else{
LOGnOUT(4,<<"*** _spSimple BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
// Note: likelihood does not improve with iterations compared to the likelihood under correction for UnObs, hence NULL
bblEM bblEM1(_tr, _sc, *_spSimple, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
,NULL,NULL); // optional &oldLnoUnObservableDataCorrection
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
newL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
//}
bblLS bbl;
MDOUBLE bestL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p); // can be sent
newL = bbl.optimizeBranches(_tr,_spVVec,_gainDist,_lossDist,_sc,_weightsUniqPatterns,_unObservableData_p,outerIter,_epsilonOptimizationBBL,_maxNumOfIterationsBBL,bestL);
if(newL<oldL){
_tr = oldTree;
LOGnOUT(4,<<"NOTE: No improvment-> Retain previous tree"<<endl);
}
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
}
// only Line-Search optimizeBranches
else{
bblLS bbl;
MDOUBLE bestL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p); // can be sent
newL = bbl.optimizeBranches(_tr,_spVVec,_gainDist,_lossDist,_sc,_weightsUniqPatterns,_unObservableData_p,outerIter,_epsilonOptimizationBBL,_maxNumOfIterationsBBL,bestL);
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
}
}
else{
if(!_isSkipBblEM){
MDOUBLE oldLnoUnObservableDataCorrection = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,NULL);
//if(isFixedRoot){
// LOGnOUT(4,<<"*** _spSimple Fix Root BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
// LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
// bblEMfixRoot bblEM1(_tr, _sc, *_spSimple, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
// ,NULL,&oldLnoUnObservableDataCorrection);
// newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
// if(_unObservableData_p)
// _unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
// newL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
//}
//else{
LOGnOUT(4,<<"*** _spSimple BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
// Note: likelihood does not improve with iterations compared to the likelihood under correction for UnObs, hence NULL
bblEM bblEM1(_tr, _sc, *_spSimple, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist
,NULL,NULL); // optional - likelihood &oldLnoUnObservableDataCorrection
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
newLnoUnObservableDataCorrection = bblEM1.getTreeLikelihood();
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
newL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
//}
}
// if include LS when BBL-EM fails
if((newL-oldL < max(minLikeImprovmentForNoLS, abs(newL/10000)) ) && _isbblLSWhenbblEMdontImprove){ // Do LS if less than 5 likelihood points were gained
LOGnOUT(4,<<" Only "<< newL-oldL<<" improvement with BBL-EM -> Perform BBL-LS one iteration"<<endl);
if(gainLossOptions::_isSkipBblEMWhenbblEMdontImprove && (newL-oldL < minLikeImprovmentForNoSkip)){
LOGnOUT(4,<<"Since no improvement (less than "<<minLikeImprovmentForNoSkip<<"), BBL-EM will be skipped next iteration, go directly to LS "<<endl);
_isSkipBblEM = true; // once BBL-EM is not improving Like, next time - skip
}
_tr = oldTree;
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
bblLS bbl;
MDOUBLE bestL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p); // can be sent
newL = bbl.optimizeBranches(_tr,_spVVec,_gainDist,_lossDist,_sc,_weightsUniqPatterns,_unObservableData_p,outerIter,_epsilonOptimizationBBL,1,bestL);
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
}
}
LOGnOUT(4,<<"*** After Branch Lengths Opt. returned Likelihood="<<"\t"<<newL<<endl);
if((newL<oldL)){
_tr = oldTree;
if(_unObservableData_p)
_unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
newL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
LOGnOUT(4,<<"NOTE: No improvment-> Retain previous tree"<<endl);
}
MDOUBLE postL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
if( !DEQUAL(newL,postL) ){
LOGnOUT(3,<<"***ERROR***: Diff returned L, and re-calculated L"<<" "<<newL<<" "<<postL<<" "<<postL-newL<<endl);
}
LOGnOUT(4,<<" Total branch lengths:"<<_tr.getAllBranchesLengthSum() <<endl);
time(&t2);
LOGnOUT(4,<<"Branch Lengths RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
return newL;
}
/********************************************************************************************
optimizeBranchLengthsvv
*********************************************************************************************/
//MDOUBLE gainLossOptimizer::optimizeBranchLengthsvv(const int outerIter)
//{
// time_t t1;
// time(&t1);
// time_t t2;
// MDOUBLE tollForPairwiseDist=0.01; // the BBL default, epsilon per branch (brent's value)
// MDOUBLE bblEMfactor = 10;
//
// int numberOfBranchs = _tr.getNodesNum();
// MDOUBLE epsilonOptimizationIterFactor = numberOfBranchs/5; // (is 1.5) for 100 branches (~50 species) the epsilon for the entire iter is 50 times the one for branch
// epsilonOptimizationIterFactor = max(5.0,epsilonOptimizationIterFactor);
// MDOUBLE epsilonOptimizationBBLIter = _epsilonOptimizationBBL*epsilonOptimizationIterFactor/bblEMfactor; // The next iteration epsilon, multiply per-branch value
// MDOUBLE oldL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
//
// MDOUBLE newL = VERYSMALL;
// tree oldTree = _tr;
// bool isFixedRoot = !_isReversible && !gainLossOptions::_isRootFreqEQstationary;
//
// if(gainLossOptions::_isBblLS){
// if (gainLossOptions::_isBblEMbeforeLSWithMissSpecifiedModel) // not implemented to work well
// {
// LOGnOUT(4,<<"*** Prior BBL-EM under stationarity assumption and one Sp\n WARN: this is not working well..."<<"\t"<<oldL<<endl);
// LOGnOUT(4,<<" BBL-EM: tollForPairwiseDist="<<tollForPairwiseDist<<" and epsilonOptimizationBBLIter="<<epsilonOptimizationBBLIter<<endl);
// if(isFixedRoot){
// LOGnOUT(4,<<"*** Start Fix Root BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
// bblEMfixRoot bblEM1(_tr, _sc, *_spSimple, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor) , epsilonOptimizationBBLIter,tollForPairwiseDist,_unObservableData_p,&oldL);
// newL = bblEM1.getTreeLikelihood();
// }
// else{
// LOGnOUT(4,<<"*** Start BBL-EM Optimization with Likelihood="<<"\t"<<oldL<<endl);
// bblEM bblEM1(_tr, _sc, *_spSimple, NULL, (int)(_maxNumOfIterationsBBL*bblEMfactor), epsilonOptimizationBBLIter,tollForPairwiseDist,_unObservableData_p,&oldL);
// newL = bblEM1.getTreeLikelihood();
// }
// LOGnOUT(4,<<"*** After BBL-EM Likelihood (with wrong model)"<<"\t"<<newL<<endl);
// if(newL<oldL){
// _tr = oldTree;
// LOGnOUT(4,<<"NOTE: No improvment-> Retain previous tree"<<endl);
// }
// }
// bblLS bbl;
// MDOUBLE bestL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p); // can be sent
// newL = bbl.optimizeBranches(_tr,_spVVec,_gainDist,_lossDist,_sc,_weightsUniqPatterns,_unObservableData_p,outerIter,_epsilonOptimizationBBL,_maxNumOfIterationsBBL,bestL);
// }
// else{
// LOGnOUT(4,<<"!!!WARNING!!!: BBL-EM is not implemented for spVVec and is not performed."<<endl);
// }
//
// LOGnOUT(4,<<"*** After Branch Lengths Opt. returned Likelihood="<<"\t"<<newL<<endl);
// if((newL<oldL)){
// _tr = oldTree;
// LOGnOUT(4,<<"NOTE: No improvment-> Retain previous tree"<<endl);
// }
// if(_unObservableData_p)
// _unObservableData_p->setLforMissingData(_tr,_spVVec,_gainDist,_lossDist);
// MDOUBLE postL = likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(_tr,_sc,_spVVec,_gainDist,_lossDist,_weightsUniqPatterns,_unObservableData_p);
//
// if( !DEQUAL(newL,postL) ){
// LOGnOUT(3,<<"***ERROR***: Diff returned L, and re-calculated L"<<" "<<newL<<" "<<postL<<" "<<postL-newL<<endl);
// }
//
// time(&t2);
// LOGnOUT(4,<<"Branch Lengths RUNNING TIME = "<<(t2-t1)/60.0<<" minutes"<<endl<<endl);
// return newL;
//}
/********************************************************************************************
*********************************************************************************************/
void gainLossOptimizer::printMixtureParams()
{
mixtureDistribution * pMixture = static_cast<mixtureDistribution*>(_sp->distr());
for (int k = 0; k < pMixture->getComponentsNum(); ++k)
{
LOGnOUT(4, << "comp="<<k<<" Alp/Beta= "<<pMixture->getAlpha(k)/pMixture->getBeta(k)<<" alpha= "<<pMixture->getAlpha(k) << " beta= " <<pMixture->getBeta(k)<<" Prob= "<<pMixture->getComponentProb(k)<<endl);
}
}
/********************************************************************************************
*********************************************************************************************/
gainLossOptions::distributionType getRateDistributionType(distribution* dist)
{
gainLossOptions::distributionType res;
if (dynamic_cast<generalGammaDistributionPlusInvariant*>(dist)){
res = gainLossOptions::GENERAL_GAMMA_PLUS_INV;
}
else if (dynamic_cast<generalGammaDistributionFixedCategories*>(dist)){
res = gainLossOptions::GENERAL_GAMMA_FIXED_CATEGORIES;
}
else if (dynamic_cast<generalGammaDistribution*>(dist)){
res = gainLossOptions::GENERAL_GAMMA;
}
else{
errorMsg::reportError("unknown type in gainLossOptions::getDistributionType");
}
return res;
}
|
