source: trunk/MagicSoft/Mars/mranforest/MRanForest.cc@ 7450

Last change on this file since 7450 was 7425, checked in by tbretz, 19 years ago
*** empty log message ***
File size: 17.6 KB
Line 
1/* ======================================================================== *\
2!
3! *
4! * This file is part of MARS, the MAGIC Analysis and Reconstruction
5! * Software. It is distributed to you in the hope that it can be a useful
6! * and timesaving tool in analysing Data of imaging Cerenkov telescopes.
7! * It is distributed WITHOUT ANY WARRANTY.
8! *
9! * Permission to use, copy, modify and distribute this software and its
10! * documentation for any purpose is hereby granted without fee,
11! * provided that the above copyright notice appear in all copies and
12! * that both that copyright notice and this permission notice appear
13! * in supporting documentation. It is provided "as is" without express
14! * or implied warranty.
15! *
16!
17!
18! Author(s): Thomas Hengstebeck 3/2003 <mailto:hengsteb@physik.hu-berlin.de>
19!
20! Copyright: MAGIC Software Development, 2000-2005
21!
22!
23\* ======================================================================== */
24
25/////////////////////////////////////////////////////////////////////////////
26//
27// MRanForest
28//
29// ParameterContainer for Forest structure
30//
31// A random forest can be grown by calling GrowForest.
32// In advance SetupGrow must be called in order to initialize arrays and
33// do some preprocessing.
34// GrowForest() provides the training data for a single tree (bootstrap
35// aggregate procedure)
36//
37// Essentially two random elements serve to provide a "random" forest,
38// namely bootstrap aggregating (which is done in GrowForest()) and random
39// split selection (which is subject to MRanTree::GrowTree())
40//
41/////////////////////////////////////////////////////////////////////////////
42#include "MRanForest.h"
43
44#include <TVector.h>
45#include <TRandom.h>
46
47#include "MHMatrix.h"
48#include "MRanTree.h"
49#include "MData.h"
50#include "MDataArray.h"
51#include "MParList.h"
52
53#include "MLog.h"
54#include "MLogManip.h"
55
56ClassImp(MRanForest);
57
58using namespace std;
59
60// --------------------------------------------------------------------------
61//
62// Default constructor.
63//
64MRanForest::MRanForest(const char *name, const char *title)
65 : fClassify(kTRUE), fNumTrees(100), fNumTry(0), fNdSize(1),
66 fRanTree(NULL), fRules(NULL), fMatrix(NULL), fUserVal(-1)
67{
68 fName = name ? name : "MRanForest";
69 fTitle = title ? title : "Storage container for Random Forest";
70
71 fForest=new TObjArray();
72 fForest->SetOwner(kTRUE);
73}
74
75MRanForest::MRanForest(const MRanForest &rf)
76{
77 // Copy constructor
78 fName = rf.fName;
79 fTitle = rf.fTitle;
80
81 fClassify = rf.fClassify;
82 fNumTrees = rf.fNumTrees;
83 fNumTry = rf.fNumTry;
84 fNdSize = rf.fNdSize;
85 fTreeNo = rf.fTreeNo;
86 fRanTree = NULL;
87
88 fRules=new MDataArray();
89 fRules->Reset();
90
91 MDataArray *newrules=rf.fRules;
92
93 for(Int_t i=0;i<newrules->GetNumEntries();i++)
94 {
95 MData &data=(*newrules)[i];
96 fRules->AddEntry(data.GetRule());
97 }
98
99 // trees
100 fForest=new TObjArray();
101 fForest->SetOwner(kTRUE);
102
103 TObjArray *newforest=rf.fForest;
104 for(Int_t i=0;i<newforest->GetEntries();i++)
105 {
106 MRanTree *rantree=(MRanTree*)newforest->At(i);
107
108 MRanTree *newtree=new MRanTree(*rantree);
109 fForest->Add(newtree);
110 }
111
112 fHadTrue = rf.fHadTrue;
113 fHadEst = rf.fHadEst;
114 fDataSort = rf.fDataSort;
115 fDataRang = rf.fDataRang;
116 fClassPop = rf.fClassPop;
117 fWeight = rf.fWeight;
118 fTreeHad = rf.fTreeHad;
119
120 fNTimesOutBag = rf.fNTimesOutBag;
121
122}
123
124// --------------------------------------------------------------------------
125// Destructor.
126MRanForest::~MRanForest()
127{
128 delete fForest;
129 if (fMatrix)
130 delete fMatrix;
131 if (fRules)
132 delete fRules;
133}
134
135void MRanForest::Print(Option_t *o) const
136{
137 *fLog << inf << GetDescriptor() << ": " << endl;
138 MRanTree *t = GetTree(0);
139 if (t)
140 {
141 *fLog << "Setting up RF for training on target:" << endl;
142 *fLog << " " << t->GetTitle() << endl;
143 }
144 if (fRules)
145 {
146 *fLog << "Following rules are used as input to RF:" << endl;
147 for (Int_t i=0;i<fRules->GetNumEntries();i++)
148 *fLog << " " << i << ") " << (*fRules)[i].GetRule() << endl;
149 }
150 *fLog << "Random forest parameters:" << endl;
151 if (t)
152 {
153 *fLog << " - " << (t->IsClassify()?"classification":"regression") << " tree" << endl;
154 *fLog << " - Number of trys: " << t->GetNumTry() << endl;
155 *fLog << " - Node size: " << t->GetNdSize() << endl;
156 }
157 *fLog << " - Number of trees: " << fNumTrees << endl;
158 *fLog << " - User value: " << fUserVal << endl;
159 *fLog << endl;
160}
161
162void MRanForest::SetNumTrees(Int_t n)
163{
164 //at least 1 tree
165 fNumTrees=TMath::Max(n,1);
166 fTreeHad.Set(fNumTrees);
167 fTreeHad.Reset();
168}
169
170void MRanForest::SetNumTry(Int_t n)
171{
172 fNumTry=TMath::Max(n,0);
173}
174
175void MRanForest::SetNdSize(Int_t n)
176{
177 fNdSize=TMath::Max(n,1);
178}
179
180void MRanForest::SetWeights(const TArrayF &weights)
181{
182 fWeight=weights;
183}
184
185void MRanForest::SetGrid(const TArrayD &grid)
186{
187 const int n=grid.GetSize();
188
189 for(int i=0;i<n-1;i++)
190 if(grid[i]>=grid[i+1])
191 {
192 *fLog<<warn<<"Grid points must be in increasing order! Ignoring grid."<<endl;
193 return;
194 }
195
196 fGrid=grid;
197
198 //*fLog<<inf<<"Following "<<n<<" grid points are used:"<<endl;
199 //for(int i=0;i<n;i++)
200 // *fLog<<inf<<" "<<i<<") "<<fGrid[i]<<endl;
201}
202
203MRanTree *MRanForest::GetTree(Int_t i) const
204{
205 return static_cast<MRanTree*>(fForest->UncheckedAt(i));
206}
207
208Int_t MRanForest::GetNumDim() const
209{
210 return fMatrix ? fMatrix->GetNcols() : 0;
211}
212
213Int_t MRanForest::GetNumData() const
214{
215 return fMatrix ? fMatrix->GetNrows() : 0;
216}
217
218Int_t MRanForest::GetNclass() const
219{
220 int maxidx = TMath::LocMax(fClass.GetSize(),fClass.GetArray());
221
222 return int(fClass[maxidx])+1;
223}
224
225void MRanForest::PrepareClasses()
226{
227 const int numdata=fHadTrue.GetSize();
228
229 if(fGrid.GetSize()>0)
230 {
231 // classes given by grid
232 const int ngrid=fGrid.GetSize();
233
234 for(int j=0;j<numdata;j++)
235 {
236 // Array is supposed to be sorted prior to this call.
237 // If match is found, function returns position of element.
238 // If no match found, function gives nearest element smaller
239 // than value.
240 int k=TMath::BinarySearch(ngrid, fGrid.GetArray(), fHadTrue[j]);
241
242 fClass[j] = k;
243 }
244
245 int minidx = TMath::LocMin(fClass.GetSize(),fClass.GetArray());
246 int min = fClass[minidx];
247 if(min!=0) for(int j=0;j<numdata;j++)fClass[j]-=min;
248
249 }else{
250 // classes directly given
251 for (Int_t j=0;j<numdata;j++)
252 fClass[j] = int(fHadTrue[j]+0.5);
253 }
254}
255
256Double_t MRanForest::CalcHadroness()
257{
258 TVector event;
259 *fRules >> event;
260
261 return CalcHadroness(event);
262}
263
264Double_t MRanForest::CalcHadroness(const TVector &event)
265{
266 Double_t hadroness=0;
267 Int_t ntree=0;
268
269 TIter Next(fForest);
270
271 MRanTree *tree;
272 while ((tree=(MRanTree*)Next()))
273 hadroness += (fTreeHad[ntree++]=tree->TreeHad(event));
274
275 return hadroness/ntree;
276}
277
278Bool_t MRanForest::AddTree(MRanTree *rantree=NULL)
279{
280 fRanTree = rantree ? rantree : fRanTree;
281
282 if (!fRanTree) return kFALSE;
283
284 MRanTree *newtree=new MRanTree(*fRanTree);
285 fForest->Add(newtree);
286
287 return kTRUE;
288}
289
290Bool_t MRanForest::SetupGrow(MHMatrix *mat,MParList *plist)
291{
292 //-------------------------------------------------------------------
293 // access matrix, copy last column (target) preliminarily
294 // into fHadTrue
295 if (fMatrix)
296 delete fMatrix;
297 fMatrix = new TMatrix(mat->GetM());
298
299 int dim = fMatrix->GetNcols()-1;
300 int numdata = fMatrix->GetNrows();
301
302 fHadTrue.Set(numdata);
303 fHadTrue.Reset(0);
304
305 for (Int_t j=0;j<numdata;j++)
306 fHadTrue[j] = (*fMatrix)(j,dim);
307
308 // remove last col
309 fMatrix->ResizeTo(numdata,dim);
310
311 //-------------------------------------------------------------------
312 // setup labels for classification/regression
313 fClass.Set(numdata);
314 fClass.Reset(0);
315
316 if (fClassify)
317 PrepareClasses();
318
319 //-------------------------------------------------------------------
320 // allocating and initializing arrays
321 fHadEst.Set(numdata);
322 fHadEst.Reset(0);
323
324 fNTimesOutBag.Set(numdata);
325 fNTimesOutBag.Reset(0);
326
327 fDataSort.Set(dim*numdata);
328 fDataSort.Reset(0);
329
330 fDataRang.Set(dim*numdata);
331 fDataRang.Reset(0);
332
333 if(fWeight.GetSize()!=numdata)
334 {
335 fWeight.Set(numdata);
336 fWeight.Reset(1.);
337 *fLog << inf <<"Setting weights to 1 (no weighting)"<< endl;
338 }
339
340 //-------------------------------------------------------------------
341 // setup rules to be used for classification/regression
342 const MDataArray *allrules=(MDataArray*)mat->GetColumns();
343 if(allrules==NULL)
344 {
345 *fLog << err <<"Rules of matrix == null, exiting"<< endl;
346 return kFALSE;
347 }
348
349 if (fRules)
350 delete fRules;
351 fRules = new MDataArray();
352 fRules->Reset();
353
354 const TString target_rule = (*allrules)[dim].GetRule();
355 for (Int_t i=0;i<dim;i++)
356 fRules->AddEntry((*allrules)[i].GetRule());
357
358 *fLog << inf << endl;
359 *fLog << "Setting up RF for training on target:" << endl;
360 *fLog << " " << target_rule.Data() << endl;
361 *fLog << "Following rules are used as input to RF:" << endl;
362 for (Int_t i=0;i<dim;i++)
363 *fLog << " " << i << ") " << (*fRules)[i].GetRule() << endl;
364 *fLog << endl;
365
366 //-------------------------------------------------------------------
367 // prepare (sort) data for fast optimization algorithm
368 if (!CreateDataSort())
369 return kFALSE;
370
371 //-------------------------------------------------------------------
372 // access and init tree container
373 fRanTree = (MRanTree*)plist->FindCreateObj("MRanTree");
374 if(!fRanTree)
375 {
376 *fLog << err << dbginf << "MRanForest, fRanTree not initialized... aborting." << endl;
377 return kFALSE;
378 }
379 //fRanTree->SetName(target_rule); // Is not stored anyhow
380
381 const Int_t tryest = TMath::Nint(TMath::Sqrt(dim));
382
383 *fLog << inf << endl;
384 *fLog << "Following input for the tree growing are used:"<<endl;
385 *fLog << " Number of Trees : "<<fNumTrees<<endl;
386 *fLog << " Number of Trials: "<<(fNumTry==0?tryest:fNumTry)<<(fNumTry==0?" (auto)":"")<<endl;
387 *fLog << " Final Node size : "<<fNdSize<<endl;
388 *fLog << " Using Grid : "<<(fGrid.GetSize()>0?"Yes":"No")<<endl;
389 *fLog << " Number of Events: "<<numdata<<endl;
390 *fLog << " Number of Params: "<<dim<<endl;
391
392 if(fNumTry==0)
393 {
394 fNumTry=tryest;
395 *fLog << inf << endl;
396 *fLog << "Set no. of trials to the recommended value of round(";
397 *fLog << TMath::Sqrt(dim) << ") = " << fNumTry << endl;
398 }
399
400 fRanTree->SetNumTry(fNumTry);
401 fRanTree->SetClassify(fClassify);
402 fRanTree->SetNdSize(fNdSize);
403
404 fTreeNo=0;
405
406 return kTRUE;
407}
408
409Bool_t MRanForest::GrowForest()
410{
411 if(!gRandom)
412 {
413 *fLog << err << dbginf << "gRandom not initialized... aborting." << endl;
414 return kFALSE;
415 }
416
417 fTreeNo++;
418
419 //-------------------------------------------------------------------
420 // initialize running output
421
422 float minfloat=fHadTrue[TMath::LocMin(fHadTrue.GetSize(),fHadTrue.GetArray())];
423 Bool_t calcResolution=(minfloat>0.001);
424
425 if (fTreeNo==1)
426 {
427 *fLog << inf << endl << underline;
428
429 if(calcResolution)
430 *fLog << "no. of tree no. of nodes resolution in % (from oob-data -> overest. of error)" << endl;
431 else
432 *fLog << "no. of tree no. of nodes rms in % (from oob-data -> overest. of error)" << endl;
433 // 12345678901234567890123456789012345678901234567890
434 }
435
436 const Int_t numdata = GetNumData();
437 const Int_t nclass = GetNclass();
438
439 //-------------------------------------------------------------------
440 // bootstrap aggregating (bagging) -> sampling with replacement:
441
442 TArrayF classpopw(nclass);
443 TArrayI jinbag(numdata); // Initialization includes filling with 0
444 TArrayF winbag(numdata); // Initialization includes filling with 0
445
446 float square=0;
447 float mean=0;
448
449 for (Int_t n=0; n<numdata; n++)
450 {
451 // The integer k is randomly (uniformly) chosen from the set
452 // {0,1,...,numdata-1}, which is the set of the index numbers of
453 // all events in the training sample
454
455 const Int_t k = Int_t(gRandom->Rndm()*numdata);
456
457 if(fClassify)
458 classpopw[fClass[k]]+=fWeight[k];
459 else
460 classpopw[0]+=fWeight[k];
461
462 mean +=fHadTrue[k]*fWeight[k];
463 square+=fHadTrue[k]*fHadTrue[k]*fWeight[k];
464
465 winbag[k]+=fWeight[k];
466 jinbag[k]=1;
467
468 }
469
470 //-------------------------------------------------------------------
471 // modifying sorted-data array for in-bag data:
472
473 // In bagging procedure ca. 2/3 of all elements in the original
474 // training sample are used to build the in-bag data
475 TArrayI datsortinbag=fDataSort;
476 Int_t ninbag=0;
477
478 ModifyDataSort(datsortinbag, ninbag, jinbag);
479
480 fRanTree->GrowTree(fMatrix,fHadTrue,fClass,datsortinbag,fDataRang,classpopw,mean,square,
481 jinbag,winbag,nclass);
482
483 //-------------------------------------------------------------------
484 // error-estimates from out-of-bag data (oob data):
485 //
486 // For a single tree the events not(!) contained in the bootstrap sample of
487 // this tree can be used to obtain estimates for the classification error of
488 // this tree.
489 // If you take a certain event, it is contained in the oob-data of 1/3 of
490 // the trees (see comment to ModifyData). This means that the classification error
491 // determined from oob-data is underestimated, but can still be taken as upper limit.
492
493 for (Int_t ievt=0;ievt<numdata;ievt++)
494 {
495 if (jinbag[ievt]>0)
496 continue;
497
498 fHadEst[ievt] +=fRanTree->TreeHad((*fMatrix), ievt);
499 fNTimesOutBag[ievt]++;
500
501 }
502
503 Int_t n=0;
504 Float_t ferr=0;
505
506 for (Int_t ievt=0;ievt<numdata;ievt++)
507 {
508 if(fNTimesOutBag[ievt]!=0)
509 {
510 float val = fHadEst[ievt]/float(fNTimesOutBag[ievt])-fHadTrue[ievt];
511 if(calcResolution) val/=fHadTrue[ievt];
512
513 ferr += val*val;
514 n++;
515 }
516 }
517 ferr = TMath::Sqrt(ferr/n);
518
519 //-------------------------------------------------------------------
520 // give running output
521 *fLog << setw(5) << fTreeNo;
522 *fLog << setw(18) << fRanTree->GetNumEndNodes();
523 *fLog << Form("%18.2f", ferr*100.);
524 *fLog << endl;
525
526 fRanTree->SetError(ferr);
527
528 // adding tree to forest
529 AddTree();
530
531 return fTreeNo<fNumTrees;
532}
533
534Bool_t MRanForest::CreateDataSort()
535{
536 // fDataSort(m,n) is the event number in which fMatrix(m,n) occurs.
537 // fDataRang(m,n) is the rang of fMatrix(m,n), i.e. if rang = r:
538 // fMatrix(m,n) is the r-th highest value of all fMatrix(m,.).
539 //
540 // There may be more then 1 event with rang r (due to bagging).
541
542 const Int_t numdata = GetNumData();
543 const Int_t dim = GetNumDim();
544
545 TArrayF v(numdata);
546 TArrayI isort(numdata);
547
548
549 for (Int_t mvar=0;mvar<dim;mvar++)
550 {
551
552 for(Int_t n=0;n<numdata;n++)
553 {
554 v[n]=(*fMatrix)(n,mvar);
555 isort[n]=n;
556
557 if(TMath::IsNaN(v[n]))
558 {
559 *fLog << err <<"Event no. "<<n<<", matrix column no. "<<mvar;
560 *fLog << err <<" has the value NaN."<<endl;
561 return kFALSE;
562 }
563 }
564
565 TMath::Sort(numdata,v.GetArray(),isort.GetArray(),kFALSE);
566
567 // this sorts the v[n] in ascending order. isort[n] is the event number
568 // of that v[n], which is the n-th from the lowest (assume the original
569 // event numbers are 0,1,...).
570
571 // control sorting
572 for(int n=1;n<numdata;n++)
573 if(v[isort[n-1]]>v[isort[n]])
574 {
575 *fLog << err <<"Event no. "<<n<<", matrix column no. "<<mvar;
576 *fLog << err <<" not at correct sorting position."<<endl;
577 return kFALSE;
578 }
579
580 for(Int_t n=0;n<numdata-1;n++)
581 {
582 const Int_t n1=isort[n];
583 const Int_t n2=isort[n+1];
584
585 fDataSort[mvar*numdata+n]=n1;
586 if(n==0) fDataRang[mvar*numdata+n1]=0;
587 if(v[n1]<v[n2])
588 {
589 fDataRang[mvar*numdata+n2]=fDataRang[mvar*numdata+n1]+1;
590 }else{
591 fDataRang[mvar*numdata+n2]=fDataRang[mvar*numdata+n1];
592 }
593 }
594 fDataSort[(mvar+1)*numdata-1]=isort[numdata-1];
595 }
596 return kTRUE;
597}
598
599void MRanForest::ModifyDataSort(TArrayI &datsortinbag, Int_t ninbag, const TArrayI &jinbag)
600{
601 const Int_t numdim=GetNumDim();
602 const Int_t numdata=GetNumData();
603
604 ninbag=0;
605 for (Int_t n=0;n<numdata;n++)
606 if(jinbag[n]==1) ninbag++;
607
608 for(Int_t m=0;m<numdim;m++)
609 {
610 Int_t k=0;
611 Int_t nt=0;
612 for(Int_t n=0;n<numdata;n++)
613 {
614 if(jinbag[datsortinbag[m*numdata+k]]==1)
615 {
616 datsortinbag[m*numdata+nt]=datsortinbag[m*numdata+k];
617 k++;
618 }else{
619 for(Int_t j=1;j<numdata-k;j++)
620 {
621 if(jinbag[datsortinbag[m*numdata+k+j]]==1)
622 {
623 datsortinbag[m*numdata+nt]=datsortinbag[m*numdata+k+j];
624 k+=j+1;
625 break;
626 }
627 }
628 }
629 nt++;
630 if(nt>=ninbag) break;
631 }
632 }
633}
634
635Bool_t MRanForest::AsciiWrite(ostream &out) const
636{
637 Int_t n=0;
638 MRanTree *tree;
639 TIter forest(fForest);
640
641 while ((tree=(MRanTree*)forest.Next()))
642 {
643 tree->AsciiWrite(out);
644 n++;
645 }
646
647 return n==fNumTrees;
648}
Note: See TracBrowser for help on using the repository browser.