]>
Commit | Line | Data |
---|---|---|
cec46807 | 1 | // --------------------------------------------------------------------------------- |
2 | // AliITStrackerANN | |
3 | // --------------------------------------------------------------------------------- | |
4 | // Neural Network-based algorithm for track reconstruction in the ALICE ITS. | |
5 | // The class contains all that is necessary in order to perform the tracking | |
6 | // using the ITS clusters in the V2 format. | |
7 | // The class is organized with some embedded objects which serve for | |
8 | // data management, and all setters and getters for all parameters. | |
9 | // The usage of the class from a tracking macro should be based essentially | |
10 | // in the initialization (constructor) and the call of a method which | |
11 | // performs in the right sequence all the operations. | |
12 | // Temporarily (and maybe definitively) all the intermediate steps are | |
13 | // public functions which could be called independently, but they do not | |
14 | // produce any result if all the preventively required steps have not been | |
15 | // performed. | |
16 | // --------------------------------------------------------------------------------- | |
17 | // Author: Alberto Pulvirenti (university of Catania) | |
18 | // Email : alberto.pulvirenti@ct.infn.it | |
19 | // --------------------------------------------------------------------------------- | |
20 | ||
cec46807 | 21 | #include <TString.h> |
22 | #include <TObjArray.h> | |
23 | #include <TVector3.h> | |
24 | #include <TFile.h> | |
25 | #include <TTree.h> | |
26 | #include <TRandom.h> | |
27 | #include <TMatrixD.h> | |
292a2409 | 28 | #if ROOT_VERSION_CODE >= ROOT_VERSION(4,0,2) |
29 | #include <TMatrixDEigen.h> | |
30 | #endif | |
cec46807 | 31 | |
32 | #include "AliITSgeom.h" | |
33 | #include "AliITStrackSA.h" | |
00a7cc50 | 34 | #include "AliITSRecPoint.h" |
cec46807 | 35 | #include "AliITStrackV2.h" |
36 | ||
37 | #include "AliITStrackerANN.h" | |
38 | ||
0db9364f | 39 | const Double_t AliITStrackerANN::fgkPi = 3.141592653; // pi |
40 | const Double_t AliITStrackerANN::fgkHalfPi = 1.570796327; // pi / 2 | |
41 | const Double_t AliITStrackerANN::fgkTwoPi = 6.283185307; // 2 * pi | |
cec46807 | 42 | |
43 | ClassImp(AliITStrackerANN) | |
44 | ||
cb729508 | 45 | |
46 | AliITStrackerANN::AliITStrackerANN() : AliITStrackerV2(), | |
47 | fVertexX(0.0), | |
48 | fVertexY(0.0), | |
49 | fVertexZ(0.0), | |
50 | fSectorNum(0), | |
51 | fSectorWidth(0), | |
52 | fPolarInterval(0), | |
53 | fCurvNum(0), | |
54 | fCurvCut(0), | |
55 | fActMinimum(0), | |
56 | fEdge1(0), | |
57 | fEdge2(0), | |
58 | fStabThreshold(0), | |
59 | fTemperature(0), | |
60 | fGain2CostRatio(0), | |
61 | fExponent(0), | |
62 | fNLayers(0), | |
63 | fFirstModInLayer(0), | |
64 | fIndexMap(0), | |
65 | fFoundTracks(0), | |
66 | fNodes(0), | |
67 | fNeurons(0), | |
68 | fMsgLevel(0), | |
69 | fStructureOK(0), | |
70 | fGeom(0){ | |
71 | //Default constructor | |
72 | } | |
cec46807 | 73 | //__________________________________________________________________________________ |
74 | AliITStrackerANN::AliITStrackerANN(const AliITSgeom *geom, Int_t msglev) | |
cb729508 | 75 | : AliITStrackerV2(geom), |
76 | fVertexX(0.0), | |
77 | fVertexY(0.0), | |
78 | fVertexZ(0.0), | |
79 | fSectorNum(0), | |
80 | fSectorWidth(0), | |
81 | fPolarInterval(0), | |
82 | fCurvNum(0), | |
83 | fCurvCut(0), | |
84 | fActMinimum(0), | |
85 | fEdge1(0), | |
86 | fEdge2(0), | |
87 | fStabThreshold(0), | |
88 | fTemperature(0), | |
89 | fGain2CostRatio(0), | |
90 | fExponent(0), | |
91 | fNLayers(0), | |
92 | fFirstModInLayer(0), | |
93 | fIndexMap(0), | |
94 | fFoundTracks(0), | |
95 | fNodes(0), | |
96 | fNeurons(0), | |
97 | fMsgLevel(msglev), | |
98 | fStructureOK(0), | |
99 | fGeom(0) | |
cec46807 | 100 | { |
101 | /************************************************************************** | |
102 | ||
103 | CONSTRUCTOR (standard-to-use version) | |
104 | ||
105 | Arguments: | |
106 | 1) the ITS geometry used in the generated event | |
107 | 2) the flag for log-messages writing | |
108 | ||
109 | The AliITSgeometry is used along the class, | |
00a7cc50 | 110 | in order to translate the local AliITSRecPoint coordinates |
cec46807 | 111 | into the Global reference frame, which is necessary for the |
112 | Neural Network algorithm to operate. | |
113 | In case of serialized use, the log messages should be excluded, | |
114 | in order to save real execution time. | |
115 | ||
116 | Operations: | |
117 | - all pointer data members are initialized | |
118 | (if possible, otherwise are set to NULL) | |
119 | - all numeric data members are initialized to | |
120 | values which allow the Neural Network to operate | |
121 | even if nothing is externally set. | |
122 | ||
123 | NOTE: it is possible that tracking an event | |
124 | with these default values results in a non-sense. | |
125 | ||
126 | **************************************************************************/ | |
127 | ||
128 | Int_t i; | |
129 | ||
130 | // Get ITS geometry | |
131 | fGeom = (AliITSgeom*)geom; | |
132 | ||
133 | // Initialize the array of first module indexes per layer | |
134 | fNLayers = geom->GetNlayers(); | |
135 | fFirstModInLayer = new Int_t[fNLayers + 1]; | |
136 | for (i = 0; i < fNLayers; i++) { | |
137 | fFirstModInLayer[i] = fGeom->GetModuleIndex(i + 1, 1, 1); | |
138 | } | |
139 | fFirstModInLayer[fNLayers] = geom->GetIndexMax(); | |
140 | ||
141 | // initialization: no curvature cut steps | |
142 | fCurvNum = 0; | |
143 | fCurvCut = 0; | |
144 | ||
145 | // initialization: 4 sectors (one for each quadrant) | |
146 | fSectorNum = 4; | |
147 | fSectorWidth = fgkHalfPi; | |
148 | ||
149 | // initialization: theta offset of 20 degrees | |
150 | fPolarInterval = 20.0; | |
151 | ||
152 | // initialization: array structure not defined | |
153 | fStructureOK = kFALSE; | |
154 | ||
155 | // initialization: vertex in the origin | |
156 | fVertexX = 0.0; | |
157 | fVertexY = 0.0; | |
158 | fVertexZ = 0.0; | |
159 | ||
160 | // initialization: uninitialized point array | |
161 | fNodes = 0; | |
162 | ||
163 | // initialization: very large (unuseful) cut values | |
164 | Int_t ilayer; | |
165 | for (ilayer = 0; ilayer < 6; ilayer++) { | |
166 | fThetaCut2D[ilayer] = TMath::Pi(); | |
167 | fThetaCut3D[ilayer] = TMath::Pi(); | |
168 | fHelixMatchCut[ilayer] = 1.0; | |
169 | } | |
170 | ||
171 | // initialization: inictial activation range between 0.3 and 0.7 | |
172 | fEdge1 = 0.3; | |
173 | fEdge2 = 0.7; | |
174 | ||
175 | // initialization: neural network operation & weights | |
176 | fTemperature = 1.0; | |
177 | fStabThreshold = 0.001; | |
178 | fGain2CostRatio = 1.0; | |
179 | fExponent = 1.0; | |
180 | fActMinimum = 0.5; | |
181 | ||
182 | // initialization: uninitialized array of neurons | |
183 | fNeurons = 0; | |
184 | ||
185 | // initialization: uninitialized array of tracks | |
186 | fFoundTracks = 0; | |
187 | } | |
188 | ||
cb729508 | 189 | |
190 | AliITStrackerANN::AliITStrackerANN(const AliITStrackerANN &n) : AliITStrackerV2((AliITStrackerV2&)n), | |
191 | fVertexX(n.fVertexX), | |
192 | fVertexY(n.fVertexY), | |
193 | fVertexZ(n.fVertexZ), | |
194 | fSectorNum(n.fSectorNum), | |
195 | fSectorWidth(n.fSectorWidth), | |
196 | fPolarInterval(n.fPolarInterval), | |
197 | fCurvNum(n.fCurvNum), | |
198 | fCurvCut(n.fCurvCut), | |
199 | fActMinimum(n.fActMinimum), | |
200 | fEdge1(n.fEdge1), | |
201 | fEdge2(n.fEdge2), | |
202 | fStabThreshold(n.fStabThreshold), | |
203 | fTemperature(n.fTemperature), | |
204 | fGain2CostRatio(n.fGain2CostRatio), | |
205 | fExponent(n.fExponent), | |
206 | fNLayers(n.fNLayers), | |
207 | fFirstModInLayer(n.fFirstModInLayer), | |
208 | fIndexMap(n.fIndexMap), | |
209 | fFoundTracks(n.fFoundTracks), | |
210 | fNodes(n.fNodes), | |
211 | fNeurons(n.fNeurons), | |
212 | fMsgLevel(n.fMsgLevel), | |
213 | fStructureOK(n.fStructureOK), | |
214 | fGeom(n.fGeom){ | |
215 | //Copy constructor | |
216 | } | |
217 | ||
218 | AliITStrackerANN& AliITStrackerANN::operator=(const AliITStrackerANN& arg){ | |
219 | //Assignment operator | |
220 | this->~AliITStrackerANN(); | |
221 | new(this) AliITStrackerANN(arg); | |
222 | return *this; | |
223 | } | |
224 | ||
cec46807 | 225 | //__________________________________________________________________________________ |
226 | void AliITStrackerANN::SetCuts | |
227 | (Int_t ncurv, Double_t *curv, Double_t *theta2D, Double_t *theta3D, Double_t *helix) | |
228 | ||
229 | /************************************************************************** | |
230 | ||
231 | CUT SETTER | |
232 | ||
233 | Allows for the definition of all kind of geometric cuts | |
234 | which have been studied in for the creation of a neuron | |
235 | from a pair of clusters C1 and C2 on consecutive layers. | |
236 | Neuron will be created only if the pair passes ALL of these cuts. | |
237 | ||
238 | At the moment, we define 4 kinds of geometrical cuts: | |
239 | a) cut on the difference of the polar 'theta' angle; | |
240 | b) cut on the angle between origin->C1 and C1->C2 in space; | |
241 | c) cut on the curvature of the circle passing | |
242 | through C1, C2 and the primary vertex; | |
243 | d) cut on heli-matching of the same three points. | |
244 | ||
245 | Arguments: | |
246 | 1) the number of curvature cut steps | |
247 | 2) the array of curvature cuts for each step | |
248 | (its dimension is given by the argument 1) | |
249 | 3) array of 5 cut values (one for each consecutive lauer pair) | |
250 | related to cut a) | |
251 | 4) array of 5 cut values (one for each consecutive lauer pair) | |
252 | related to cut b) | |
253 | 5) array of 5 cut values (one for each consecutive lauer pair) | |
254 | related to cut c) | |
255 | ||
256 | Operations: | |
257 | - gets the values for each cut and stores them in data members | |
258 | - in the case of curvature cuts, the cut array | |
259 | (whose size is not fixed) is allocated | |
260 | ||
261 | NOTE: in case that the user wants to set onyl ONE of the 4 cuts array, | |
262 | he can simply pass NULL arguments for other cuts, and (eventually) | |
263 | a ZERO as the first argument (if curvature cuts have not to be set). | |
264 | ||
265 | Anyway, all the cuts have to be set at least once. | |
266 | ||
267 | **************************************************************************/ | |
268 | { | |
269 | // counter | |
270 | Int_t i; | |
271 | ||
272 | /*** Curvature cut setting ***/ | |
273 | ||
274 | // first of all, the curvature cuts are sorted in increasing order | |
275 | // (from the smallest to the largest one) | |
276 | Int_t *ind = new Int_t[ncurv]; | |
277 | TMath::Sort(ncurv, curv, ind, kFALSE); | |
278 | // then, the curvature cut array is allocated and filled | |
279 | // (a message with the list of defined cuts can be optionally shown) | |
280 | fCurvCut = new Double_t[ncurv]; | |
281 | if (fMsgLevel >= 1) cout << "Number of curvature cuts: " << ncurv << endl; | |
282 | for (i = 0; i < ncurv; i++) { | |
283 | fCurvCut[i] = curv[ind[i]]; | |
284 | if (fMsgLevel >= 1) cout << " - " << fCurvCut[i] << endl; | |
285 | } | |
286 | fCurvNum = ncurv; | |
287 | ||
288 | /*** 'Fixed' cuts setting ***/ | |
289 | ||
290 | // checks what cuts have to be set | |
291 | Bool_t doTheta2D = (theta2D != 0); | |
292 | Bool_t doTheta3D = (theta3D != 0); | |
293 | Bool_t doHelix = (helix != 0); | |
294 | // sets the cuts for all layer pairs | |
295 | for (i = 0; i < fNLayers - 1; i++) { | |
296 | if (doTheta2D) fThetaCut2D[i] = theta2D[i]; | |
297 | if (doTheta3D) fThetaCut3D[i] = theta3D[i]; | |
298 | if (doHelix) fHelixMatchCut[i] = helix[i]; | |
299 | } | |
300 | // if required, lists the cuts | |
301 | if (!fMsgLevel < 2) return; | |
302 | cout << "Theta 2D cuts: "; | |
303 | if (!doTheta2D) { | |
304 | cout << "<not set>" << endl; | |
305 | } | |
306 | else { | |
307 | cout << endl; | |
308 | for (i = 0; i < fNLayers - 1; i++) { | |
309 | cout << "For layers " << i << " --> " << i + 1; | |
310 | cout << " cut = " << fThetaCut2D[i] << endl; | |
311 | } | |
312 | } | |
313 | cout << "---" << endl; | |
314 | cout << "Theta 3D cuts: "; | |
315 | if (!doTheta3D) { | |
316 | cout << "<not set>" << endl; | |
317 | } | |
318 | else { | |
319 | cout << endl; | |
320 | for (i = 0; i < fNLayers - 1; i++) { | |
321 | cout << "For layers " << i << " --> " << i + 1; | |
322 | cout << " cut = " << fThetaCut3D[i] << endl; | |
323 | } | |
324 | } | |
325 | cout << "---" << endl; | |
326 | cout << "Helix-match cuts: "; | |
327 | if (!doHelix) { | |
328 | cout << "<not set>" << endl; | |
329 | } | |
330 | else { | |
331 | cout << endl; | |
332 | for (i = 0; i < fNLayers - 1; i++) { | |
333 | cout << "For layers " << i << " --> " << i + 1; | |
334 | cout << " cut = " << fHelixMatchCut[i] << endl; | |
335 | } | |
336 | } | |
337 | cout << "---" << endl; | |
338 | } | |
339 | ||
340 | //__________________________________________________________________________________ | |
341 | Bool_t AliITStrackerANN::GetGlobalXYZ | |
342 | (Int_t refIndex, | |
343 | Double_t &x, Double_t &y, Double_t &z, Double_t &ex2, Double_t &ey2, Double_t &ez2) | |
344 | ||
345 | /************************************************************************** | |
346 | ||
347 | LOCAL TO GLOBAL TRANSLATOR | |
348 | ||
349 | Taking information from the ITS geometry stored in the class, | |
350 | gets a stored AliITScluster and calculates it coordinates | |
351 | and errors in the global reference frame. | |
352 | These values are stored in the variables, | |
353 | which are passed by reference. | |
354 | ||
355 | Arguments: | |
356 | 1) reference index for the cluster to use | |
357 | 2) (by reference) place to store the global X-coord into | |
358 | 3) (by reference) place to store the global Y-coord into | |
359 | 4) (by reference) place to store the global Z-coord into | |
360 | 5) (by reference) place to store the global X-coord error into | |
361 | 6) (by reference) place to store the global Y-coord error into | |
362 | 7) (by reference) place to store the global Z-coord error into | |
363 | ||
364 | Operations: | |
365 | essentially, determines the ITS module index from the | |
00a7cc50 | 366 | detector index of the AliITSRecPoint object, and extracts |
cec46807 | 367 | the roto-translation from the ITS geometry, to convert |
368 | the local module coordinates into the global ones. | |
369 | ||
370 | Return value: | |
371 | - kFALSE if the given cluster index points to a non-existing | |
372 | cluster, or if the layer number makes no sense (< 0 or > 6). | |
373 | - otherwise, kTRUE (meaning a successful operation). | |
374 | ||
375 | **************************************************************************/ | |
376 | { | |
377 | // checks if the layer is correct | |
378 | Int_t ilayer = (refIndex & 0xf0000000) >> 28; | |
379 | if (ilayer < 0 || ilayer >= fNLayers) { | |
380 | Error("GetGlobalXYZ", "Wrong layer number: %d [range: %d - %d]", ilayer, 0, fNLayers); | |
381 | return kFALSE; | |
382 | } | |
383 | // checks if the referenced cluster exists and corresponds to the passed reference | |
00a7cc50 | 384 | AliITSRecPoint *refCluster = (AliITSRecPoint*) GetCluster(refIndex); |
cec46807 | 385 | if (!refCluster) { |
386 | Error("GetGlobalXYZ", "Cluster not found for index %d", refIndex); | |
387 | return kFALSE; | |
388 | } | |
389 | ||
390 | // determine the detector number | |
391 | Int_t detID = refCluster->GetDetectorIndex() + fFirstModInLayer[ilayer]; | |
392 | ||
393 | // get rotation matrix | |
394 | Double_t rot[9]; | |
395 | fGeom->GetRotMatrix(detID, rot); | |
396 | ||
397 | // get translation vector | |
398 | Float_t tx,ty,tz; | |
399 | fGeom->GetTrans(detID, tx, ty, tz); | |
400 | ||
401 | // determine r and phi for the reference conversion | |
402 | Double_t r = -(Double_t)tx * rot[1] + (Double_t)ty * rot[0]; | |
403 | if (ilayer == 0) r = -r; | |
404 | Double_t phi = TMath::ATan2(rot[1],rot[0]); | |
405 | if (ilayer == 0) phi -= fgkPi; | |
406 | ||
407 | // sets values for X, Y, Z in global coordinates and their errors | |
408 | Double_t cosPhi = TMath::Cos(phi); | |
409 | Double_t sinPhi = TMath::Sin(phi); | |
410 | x = r*cosPhi + refCluster->GetY()*sinPhi; | |
411 | y = -r*sinPhi + refCluster->GetY()*cosPhi; | |
412 | z = refCluster->GetZ(); | |
413 | ex2 = refCluster->GetSigmaY2()*sinPhi*sinPhi; | |
414 | ey2 = refCluster->GetSigmaY2()*cosPhi*cosPhi; | |
415 | ez2 = refCluster->GetSigmaZ2(); | |
416 | ||
417 | return kTRUE; | |
418 | } | |
419 | ||
420 | //__________________________________________________________________________________ | |
421 | AliITStrackerANN::AliITSnode* AliITStrackerANN::AddNode(Int_t refIndex) | |
422 | ||
423 | /************************************************************************** | |
424 | ||
425 | GENERATOR OF NEURAL NODES | |
426 | ||
427 | Fills the array of neural 'nodes', which are the ITS clusters | |
428 | translated in the global reference frame. | |
429 | Given that the global coordinates are used many times, they are | |
430 | stored in a well-defined structure, in the form of an embedded class. | |
431 | Moreover, this class allows a faster navigation among points | |
432 | and neurons, by means of some object arrays, storing only the | |
433 | neurons which start from, or end to, the given node. | |
434 | Finally, each node contains all the other nodes which match it | |
435 | with respect to the fixed walues, in order to perform a faster | |
436 | neuron-creation phase. | |
437 | ||
438 | Arguments: | |
00a7cc50 | 439 | 1) reference index of the correlated AliITSRecPoint object |
cec46807 | 440 | |
441 | Operations: | |
442 | - allocates the new AliITSnode objects | |
443 | - initializes its object arrays | |
444 | - from the global coordinates, calculates the | |
445 | 'phi' and 'theta' coordinates, in order to store it | |
446 | into the correct theta-slot and azimutal sector. | |
447 | ||
448 | REturn values: | |
449 | - the pointer of the creater AliITSnode object | |
450 | - in case of errors, a waring is given and a NULL is returned | |
451 | ||
452 | **************************************************************************/ | |
453 | { | |
454 | // create object and set the reference | |
455 | AliITSnode *node = new AliITSnode; | |
456 | if (!node) { | |
457 | Warning("AddNode", "Error occurred when allocating AliITSnode"); | |
458 | return 0; | |
459 | } | |
460 | node->ClusterRef() = refIndex; | |
461 | ||
462 | // calls the conversion function, which makes also some checks | |
463 | // (layer number within range, existence of referenced cluster) | |
464 | if ( !GetGlobalXYZ ( | |
465 | refIndex, | |
466 | node->X(), node->Y(), node->Z(), | |
467 | node->ErrX2(), node->ErrY2(), node->ErrZ2() | |
468 | ) ) {return 0;} | |
469 | ||
470 | // initializes the object arrays | |
471 | node->Matches() = new TObjArray; | |
472 | node->InnerOf() = new TObjArray; | |
473 | node->OuterOf() = new TObjArray; | |
474 | ||
475 | // finds azimutal and polar sector (in degrees) | |
476 | Double_t phi = node->GetPhi() * 180.0 / fgkPi; | |
477 | Double_t theta = node->GetTheta() * 180.0 / fgkPi; | |
478 | Int_t isector = (Int_t)(phi / fSectorWidth); | |
479 | Int_t itheta = (Int_t)theta; | |
480 | Int_t ilayer = (refIndex & 0xf0000000) >> 28; | |
481 | ||
482 | // selects the right TObjArray to store object into | |
483 | TObjArray *sector = (TObjArray*)fNodes[ilayer][itheta]->At(isector); | |
484 | sector->AddLast(node); | |
485 | ||
486 | return node; | |
487 | } | |
488 | ||
489 | //__________________________________________________________________________________ | |
490 | void AliITStrackerANN::CreateArrayStructure(Int_t nsectors) | |
491 | ||
492 | /************************************************************************** | |
493 | ||
494 | ARRAY STRUCTURE CREATOR | |
495 | ||
496 | Creates a structure of nested TObjArray's where the AliITSnode's | |
497 | have to be stored: | |
498 | - the first level is made by 6 arrays (one for each layer) | |
499 | - the second level is made by 180 arrays (one for each int part of theta) | |
500 | - the third level is made by a variable number of arrays | |
501 | (one for each azimutal sector) | |
502 | ||
503 | Arguments: | |
504 | 1) the number of azimutal sectors | |
505 | ||
506 | Operations: | |
507 | - calculates the width of each sector, from the argument | |
508 | - allocates and initializes all array levels | |
509 | - sets a flag which tells the user if this NECESSARY operation | |
510 | has been performed (it is needed BEFORE performing tracking) | |
511 | ||
512 | **************************************************************************/ | |
513 | { | |
514 | // Set the number of sectors and their width. | |
515 | fSectorNum = nsectors; | |
516 | fSectorWidth = 360.0 / (Double_t)fSectorNum; | |
517 | if (fMsgLevel >= 2) { | |
518 | cout << fSectorNum << " sectors --> sector width (degrees) = " << fSectorWidth << endl; | |
519 | } | |
520 | ||
521 | // Meaningful indexes | |
522 | Int_t ilayer, isector, itheta; | |
523 | ||
524 | // Mark for the created objects | |
525 | TObjArray *sector = 0; | |
526 | ||
527 | // First index: layer | |
528 | fNodes = new TObjArray**[fNLayers]; | |
529 | for (ilayer = 0; ilayer < fNLayers; ilayer++) { | |
530 | fNodes[ilayer] = new TObjArray*[180]; | |
531 | for (itheta = 0; itheta < 180; itheta++) fNodes[ilayer][itheta] = 0; | |
532 | for (itheta = 0; itheta < 180; itheta++) { | |
533 | fNodes[ilayer][itheta] = new TObjArray(nsectors); | |
534 | for (isector = 0; isector < nsectors; isector++) { | |
535 | sector = new TObjArray; | |
536 | sector->SetOwner(); | |
537 | fNodes[ilayer][itheta]->AddAt(sector, isector); | |
538 | } | |
539 | } | |
540 | } | |
541 | ||
542 | // Sets a checking flag to TRUE. | |
543 | // This flag is checked before filling up the arrays with the points. | |
544 | fStructureOK = kTRUE; | |
545 | } | |
546 | ||
547 | //__________________________________________________________________________________ | |
548 | Int_t AliITStrackerANN::ArrangePoints(char *exportFile) | |
549 | ||
550 | /************************************************************************** | |
551 | ||
552 | POINTS LOCATOR | |
553 | ||
554 | This function assembles the operation from the other above methods, | |
555 | and fills the arrays with the clusters already stored in the | |
556 | layers of the tracker. | |
557 | Then, in order to use this method, the user MUSTs call LoadClusters() | |
558 | before. | |
559 | ||
560 | Arguments: | |
561 | 1) string for a file name where the global ccordinates | |
562 | of all points can be exported (optional). | |
563 | If this file must not be created, simply pass a NULL argument | |
564 | ||
565 | Operations: | |
00a7cc50 | 566 | - for each AliITSRecPoint in each AliITSlayer, a ne AliITSnode |
cec46807 | 567 | is created and stored in the correct location. |
568 | ||
569 | Return values: | |
570 | - the number of stored points | |
571 | - when errors occur, or no points are found, 0 is returned | |
572 | ||
573 | **************************************************************************/ | |
574 | { | |
575 | // Check if the array structure has been created | |
576 | if (!fStructureOK) { | |
577 | Error("ArrangePoints", "Structure NOT defined. Call CreateArrayStructure() first"); | |
578 | return 0; | |
579 | } | |
580 | ||
581 | // meaningful indexes | |
582 | Int_t ientry, ilayer, nentries = 0, index; | |
583 | Int_t nPtsLayer = 0; | |
584 | ||
585 | // if the argument is not NULL, a file is opened | |
586 | fstream file(exportFile, ios::out); | |
587 | if (!exportFile || file.fail()) { | |
588 | file.close(); | |
589 | exportFile = 0; | |
590 | } | |
591 | ||
592 | // scan all layers for node creation | |
593 | for (ilayer = 0; ilayer < fNLayers; ilayer++) { | |
594 | nPtsLayer = GetNumberOfClustersLayer(ilayer); | |
595 | if (fMsgLevel >= 1) { | |
596 | cout << "Layer " << ilayer << " --> " << nPtsLayer << " clusters" << endl; | |
597 | } | |
598 | for (ientry = 0; ientry < nPtsLayer; ientry++) { | |
599 | // calculation of cluster index : (Bit mask LLLLIIIIIIIIIIII) | |
600 | // (L = bits used for layer) | |
601 | // (I = bits used for position in layer) | |
602 | index = ilayer << 28; | |
603 | index += ientry; | |
604 | // add new AliITSnode object | |
605 | AliITSnode *n = AddNode(index); | |
606 | if ( (n != NULL) && exportFile ) { | |
607 | file << index << ' ' << n->X() << ' ' << n->Y() << ' ' << n->Z() << endl; | |
608 | } | |
609 | } | |
610 | nentries += nPtsLayer; | |
611 | } | |
612 | ||
613 | // a conventional final message is put at the end of file | |
614 | if (exportFile) { | |
615 | file << "-1 0.0 0.0 0.0" << endl; | |
616 | file.close(); | |
617 | } | |
618 | ||
619 | // returns the number of points processed | |
620 | return nentries; | |
621 | } | |
622 | ||
623 | //__________________________________________________________________________________ | |
624 | void AliITStrackerANN::StoreOverallMatches() | |
625 | ||
626 | /************************************************************************** | |
627 | ||
628 | NODE-MATCH ANALYSIS | |
629 | ||
630 | Once the nodes have been created, a firs analysis is to check | |
631 | what pairs will satisfy at least the 'fixed' cuts (theta, helix-match) | |
632 | and the most permissive (= larger) curvature cut. | |
633 | All these node pairs are suitable for neuron creation. | |
634 | In fact, when performing a Neural Tracking step, the only further check | |
635 | will be a check against the current curvature step, while the other | |
636 | are always the same. | |
637 | For thi purpose, each AliITSnode has a data member, named 'fMatches' | |
638 | which contains references to all other AliITSnodes in the successive layer | |
639 | that form, with it, a 'good' pair, with respect to the above cited cuts. | |
640 | Then, in each step for neuron creation, the possible neurons starting from | |
641 | each node will be searched ONLY within the nodes referenced in fMatches. | |
642 | This, of course, speeds up a lot the neuron creation procedure, at the | |
643 | cost of some memory occupation, which results not to be critical. | |
644 | ||
645 | Operations: | |
646 | - for each AliITSnode, matches are found according to the criteria | |
647 | expressed above, and stored in the node->fMatches array | |
648 | ||
649 | **************************************************************************/ | |
650 | { | |
651 | // meaningful counters | |
652 | Int_t ilayer, isector, itheta1, itheta2, check; | |
653 | TObjArray *list1 = 0, *list2 = 0; | |
654 | AliITSnode *node1 = 0, *node2 = 0; | |
655 | Double_t thetaMin, thetaMax; | |
656 | Int_t imin, imax; | |
657 | ||
658 | // Scan for each sector | |
659 | for (isector = 0; isector < fSectorNum; isector++) { | |
660 | // sector is chosen once for both lists | |
661 | for (ilayer = 0; ilayer < fNLayers - 1; ilayer++) { | |
662 | for (itheta1 = 0; itheta1 < 180; itheta1++) { | |
663 | list1 = (TObjArray*)fNodes[ilayer][itheta1]->At(isector); | |
664 | TObjArrayIter iter1(list1); | |
665 | while ( (node1 = (AliITSnode*)iter1.Next()) ) { | |
666 | if (node1->IsUsed()) continue; | |
667 | // clear an eventually already present array | |
668 | // node1->Matches()->Clear(); | |
669 | // get the global coordinates and defines the theta interval from cut | |
670 | thetaMin = (node1->GetTheta() * 180.0 / fgkPi) - fPolarInterval; | |
671 | thetaMax = (node1->GetTheta() * 180.0 / fgkPi) + fPolarInterval; | |
672 | imin = (Int_t)thetaMin; | |
673 | imax = (Int_t)thetaMax; | |
674 | if (imin < 0) imin = 0; | |
675 | if (imax > 179) imax = 179; | |
676 | // loop on the selected theta slots | |
677 | for (itheta2 = imin; itheta2 <= imax; itheta2++) { | |
678 | list2 = (TObjArray*)fNodes[ilayer + 1][itheta2]->At(isector); | |
679 | TObjArrayIter iter2(list2); | |
680 | while ( (node2 = (AliITSnode*)iter2.Next()) ) { | |
681 | check = PassAllCuts(node1, node2, fCurvNum - 1, fVertexX, fVertexY, fVertexZ); | |
682 | if (check == 0) { | |
683 | node1->Matches()->AddLast(node2); | |
684 | } | |
685 | } // while (node2...) | |
686 | } // for (itheta2...) | |
687 | } // while (node1...) | |
688 | } // for (itheta...) | |
689 | } // for (ilayer...) | |
690 | } // for (isector...) | |
691 | } | |
692 | ||
693 | //__________________________________________________________________________________ | |
694 | Int_t AliITStrackerANN::PassAllCuts | |
695 | (AliITSnode *inner, AliITSnode *outer, Int_t curvStep, | |
696 | Double_t vx, Double_t vy, Double_t vz) | |
697 | { | |
698 | // *********************************************************************************** | |
699 | // | |
700 | // This check is called in the above method for finding the matches of each node | |
701 | // It check the passed point pair against all the fixed cuts and a specified | |
702 | // curvature cut, among all the ones which have been defined. | |
703 | // The cuts need a vertex-constraint, which is not absolute, but it is passed | |
704 | // as argument. | |
705 | // | |
706 | // Arguments: | |
707 | // 1) the point in the inner layer | |
708 | // 2) the point in the outer layer | |
709 | // 3) curvature step for the curvature cut check (preferably the last) | |
710 | // 4) X of the used vertex | |
711 | // 5) Y of the used vertex | |
712 | // 6) Z of the used vertex | |
713 | // | |
714 | // Operations: | |
715 | // - if necessary, swaps the two points | |
716 | // (the first must be in the innermost of the two layers) | |
717 | // - checks for the theta cuts | |
718 | // - calculates the circle passing through the vertex | |
719 | // and the given points and checks for the curvature cut | |
720 | // - using the radius calculated there, checks for the helix-math cut | |
721 | // | |
722 | // Return values: | |
723 | // 0 - All cuts passed | |
724 | // 1 - theta 2D cut not passed | |
725 | // 2 - theta 3D cut not passed | |
726 | // 3 - curvature calculated but cut not passed | |
727 | // 4 - curvature not calculated (division by zero) | |
728 | // 5 - helix cut not passed | |
729 | // 6 - curvature inxed out of range | |
730 | // | |
731 | // *********************************************************************************** | |
732 | ||
733 | // Check for curvature index | |
734 | if (curvStep < 0 || curvStep >= fCurvNum) return 6; | |
735 | ||
736 | // Swap points in order that r1 < r2 | |
737 | AliITSnode *temp = 0; | |
738 | if (outer->GetLayer() < inner->GetLayer()) { | |
739 | temp = outer; | |
740 | outer = inner; | |
741 | inner = temp; | |
742 | } | |
743 | ||
744 | // All cuts are variable according to the layer of the | |
745 | // innermost point (the other point will surely be | |
746 | // in the further one, because we don't check poin pairs | |
747 | // which are not in adjacent layers) | |
748 | // The reference is given by the innermost point. | |
749 | Int_t layRef = inner->GetLayer(); | |
750 | ||
751 | // The calculations in the transverse plane are made in | |
752 | // a shifted reference frame, whose origin corresponds to | |
753 | // the reference point passed in the argument. | |
754 | Double_t xIn = inner->X() - vx; | |
755 | Double_t xOut = outer->X() - vx; | |
756 | Double_t yIn = inner->Y() - vy; | |
757 | Double_t yOut = outer->Y() - vy; | |
758 | Double_t zIn = inner->Z() - vz; | |
759 | Double_t zOut = outer->Z() - vz; | |
760 | Double_t rIn = TMath::Sqrt(xIn*xIn + yIn*yIn); | |
761 | Double_t rOut = TMath::Sqrt(xOut*xOut + yOut*yOut); | |
762 | ||
763 | // Check for theta cut. | |
764 | // There are two different angular cuts: | |
765 | // one w.r. to the angle in the 2-dimensional r-z plane... | |
766 | Double_t dthetaRZ; | |
767 | TVector3 origin2innerRZ(zIn, rIn, 0.0); | |
768 | TVector3 inner2outerRZ(zOut - zIn, rOut - rIn, 0.0); | |
769 | dthetaRZ = origin2innerRZ.Angle(inner2outerRZ) * 180.0 / fgkPi; | |
770 | if (dthetaRZ > fThetaCut2D[layRef]) { | |
771 | return 1; | |
772 | // r-z theta cut not passed ---> 1 | |
773 | } | |
774 | // ...and another w.r. to the angle in the 3-dimensional x-y-z space | |
775 | Double_t dthetaXYZ; | |
776 | TVector3 origin2innerXYZ(xIn, yIn, zIn); | |
777 | TVector3 inner2outerXYZ(xOut - xIn, yOut - yIn, zOut - zIn); | |
778 | dthetaXYZ = origin2innerXYZ.Angle(inner2outerXYZ) * 180.0 / fgkPi; | |
779 | if (dthetaXYZ > fThetaCut3D[layRef]) { | |
780 | return 2; | |
781 | // x-y-z theta cut not passed ---> 2 | |
782 | } | |
783 | ||
784 | // Calculation & check of curvature | |
785 | Double_t dx = xIn - xOut; | |
786 | Double_t dy = yIn - yOut; | |
787 | Double_t num = 2.0 * (xIn*yOut - xOut*yIn); | |
788 | Double_t den = rIn*rOut*sqrt(dx*dx + dy*dy); | |
789 | Double_t curv = 0.; | |
790 | if (den != 0.) { | |
791 | curv = TMath::Abs(num / den); | |
792 | if (curv > fCurvCut[curvStep]) { | |
793 | return 3; | |
794 | // curvature too large for cut ---> 3 | |
795 | } | |
796 | } | |
797 | else { | |
798 | Error("PassAllCuts", "Curvature calculations gives zero denominator"); | |
799 | return 4; | |
800 | // error: denominator = 0 ---> 4 | |
801 | } | |
802 | ||
803 | // Calculation & check of helix matching | |
804 | Double_t helMatch = 0.0; | |
805 | Double_t arcIn = 2.0 * rIn * curv; | |
806 | Double_t arcOut = 2.0 * rOut * curv; | |
807 | if (arcIn > -1.0 && arcIn < 1.0) | |
808 | arcIn = TMath::ASin(arcIn); | |
809 | else | |
810 | arcIn = ((arcIn > 0.0) ? 0.5 : 1.5) * TMath::Pi(); | |
811 | if (arcOut > -1.0 && arcOut < 1.0) | |
812 | arcOut = TMath::ASin(arcOut); | |
813 | else | |
814 | arcOut = ((arcOut > 0.0) ? 0.5 : 1.5) * TMath::Pi(); | |
815 | arcIn /= 2.0 * curv; | |
816 | arcOut /= 2.0 * curv; | |
817 | if (arcIn == 0.0 || arcOut == 0.0) { | |
818 | Error("PassAllCuts", "Calculation returns zero-length arcs: l1=%f, l2=%f", arcIn, arcOut); | |
819 | return 4; | |
820 | // error: circumference arcs seem to equal zero ---> 4 | |
821 | } | |
822 | helMatch = TMath::Abs(zIn / arcIn - zOut / arcOut); | |
823 | if (helMatch > fHelixMatchCut[layRef]) { | |
824 | return 5; | |
825 | // helix match cut not passed ---> 5 | |
826 | } | |
827 | ||
828 | // ALL CUTS PASSED ---> 0 | |
829 | return 0; | |
830 | } | |
831 | ||
832 | //__________________________________________________________________________________ | |
833 | Bool_t AliITStrackerANN::PassCurvCut | |
834 | (AliITSnode *inner, AliITSnode *outer, | |
835 | Int_t curvStep, | |
836 | Double_t vx, Double_t vy, Double_t vz) | |
837 | { | |
838 | //*********************************************************************************** | |
839 | // | |
840 | // This method operates essentially like the above one, but it is used | |
841 | // during a single step of Neural Tracking, where the curvature cut | |
842 | // changes. | |
843 | // Then, not necessaryly all the nodes stored in the fMatches array | |
844 | // will be suitable for neuron creation in an intermediate step. | |
845 | // | |
846 | // It has the same arguments of the PassAllCuts() method, but | |
847 | // the theta cut is not checked. | |
848 | // Moreover, it has a boolean return value. | |
849 | // | |
850 | //*********************************************************************************** | |
851 | ||
852 | // Check for curvature index | |
853 | if (curvStep < 0 || curvStep >= fCurvNum) return 6; | |
854 | ||
855 | // Find the reference layer | |
856 | Int_t layIn = inner->GetLayer(); | |
857 | Int_t layOut = outer->GetLayer(); | |
858 | Int_t layRef = (layIn < layOut) ? layIn : layOut; | |
859 | ||
860 | // The calculations in the transverse plane are made in | |
861 | // a shifted reference frame, whose origin corresponds to | |
862 | // the reference point passed in the argument. | |
863 | Double_t xIn = inner->X() - vx; | |
864 | Double_t xOut = outer->X() - vx; | |
865 | Double_t yIn = inner->Y() - vy; | |
866 | Double_t yOut = outer->Y() - vy; | |
867 | Double_t zIn = inner->Z() - vz; | |
868 | Double_t zOut = outer->Z() - vz; | |
869 | Double_t rIn = TMath::Sqrt(xIn*xIn + yIn*yIn); | |
870 | Double_t rOut = TMath::Sqrt(xOut*xOut + yOut*yOut); | |
871 | ||
872 | // Calculation & check of curvature | |
873 | Double_t dx = xIn - xOut; | |
874 | Double_t dy = yIn - yOut; | |
875 | Double_t num = 2.0 * (xIn*yOut - xOut*yIn); | |
876 | Double_t den = rIn*rOut*sqrt(dx*dx + dy*dy); | |
877 | Double_t curv = 0.; | |
878 | /* OLD VERSION | |
879 | if (den != 0.) { | |
880 | curv = TMath::Abs(num / den); | |
881 | if (curv > fCurvCut[curvStep]) return kFALSE; | |
882 | return kTRUE; | |
883 | } | |
884 | else | |
885 | return kFALSE; | |
886 | */ | |
887 | // NEW VERSION | |
888 | if (den != 0.) { | |
889 | curv = TMath::Abs(num / den); | |
890 | if (curv > fCurvCut[curvStep]) { | |
891 | return kFALSE; | |
892 | } | |
893 | } | |
894 | else { | |
895 | Error("PassAllCuts", "Curvature calculations gives zero denominator"); | |
896 | return kFALSE; | |
897 | } | |
898 | ||
899 | // Calculation & check of helix matching | |
900 | Double_t helMatch = 0.0; | |
901 | Double_t arcIn = 2.0 * rIn * curv; | |
902 | Double_t arcOut = 2.0 * rOut * curv; | |
903 | if (arcIn > -1.0 && arcIn < 1.0) | |
904 | arcIn = TMath::ASin(arcIn); | |
905 | else | |
906 | arcIn = ((arcIn > 0.0) ? 0.5 : 1.5) * TMath::Pi(); | |
907 | if (arcOut > -1.0 && arcOut < 1.0) | |
908 | arcOut = TMath::ASin(arcOut); | |
909 | else | |
910 | arcOut = ((arcOut > 0.0) ? 0.5 : 1.5) * TMath::Pi(); | |
911 | arcIn /= 2.0 * curv; | |
912 | arcOut /= 2.0 * curv; | |
913 | if (arcIn == 0.0 || arcOut == 0.0) { | |
914 | Error("PassAllCuts", "Calculation returns zero-length arcs: l1=%f, l2=%f", arcIn, arcOut); | |
915 | return 4; | |
916 | // error: circumference arcs seem to equal zero ---> 4 | |
917 | } | |
918 | helMatch = TMath::Abs(zIn / arcIn - zOut / arcOut); | |
919 | return (helMatch <= fHelixMatchCut[layRef]); | |
920 | } | |
921 | ||
922 | //__________________________________________________________________________________ | |
923 | void AliITStrackerANN::PrintMatches(Bool_t stop) | |
924 | { | |
925 | // Prints the list of points which appear to match | |
926 | // each one of them, according to the preliminary | |
927 | // overall cuts. | |
928 | // The arguments states if a pause is required after printing | |
929 | // the matches for each one. In this case, a keypress is needed. | |
930 | ||
931 | TObjArray *sector = 0; | |
932 | Int_t ilayer, isector, itheta, nF; | |
933 | AliITSnode *node1 = 0, *node2 = 0; | |
00a7cc50 | 934 | //AliITSRecPoint *cluster1 = 0, *cluster2 = 0; |
cec46807 | 935 | |
936 | for (ilayer = 0; ilayer < 6; ilayer++) { | |
937 | for (isector = 0; isector < fSectorNum; isector++) { | |
938 | for (itheta = 0; itheta < 180; itheta++) { | |
939 | sector = (TObjArray*)fNodes[ilayer][itheta]->At(isector); | |
940 | TObjArrayIter points(sector); | |
941 | while ( (node1 = (AliITSnode*)points.Next()) ) { | |
942 | nF = (Int_t)node1->Matches()->GetEntries(); | |
943 | cout << "Node layer: " << node1->GetLayer() << " --> "; | |
944 | if (!nF) { | |
945 | cout << "NO Matches!!!" << endl; | |
946 | continue; | |
947 | } | |
948 | cout << nF << " Matches" << endl; | |
949 | cout << "Reference cluster: " << hex << node1->ClusterRef() << endl; | |
950 | TObjArrayIter matches(node1->Matches()); | |
951 | while ( (node2 = (AliITSnode*)matches.Next()) ) { | |
952 | cout << "Match with " << hex << node2->ClusterRef() << endl; | |
953 | } | |
954 | if (stop) { | |
955 | cout << "Press a key" << endl; | |
956 | cin.get(); | |
957 | } | |
958 | } | |
959 | } | |
960 | } | |
961 | } | |
962 | } | |
963 | ||
964 | //__________________________________________________________________________________ | |
965 | void AliITStrackerANN::ResetNodes(Int_t isector) | |
966 | { | |
967 | /*********************************************************************************** | |
968 | ||
969 | NODE NEURON ARRAY CLEANER | |
970 | ||
971 | After a neural tracking step, this method | |
972 | clears the arrays 'fInnerOf' and 'fOuterOf' of each AliITSnode | |
973 | ||
974 | Arguments: | |
975 | - the sector where the operation is being executed | |
976 | ||
977 | ***********************************************************************************/ | |
978 | ||
979 | Int_t ilayer, itheta; | |
980 | TObjArray *sector = 0; | |
981 | AliITSnode *node = 0; | |
982 | for (ilayer = 0; ilayer < fNLayers; ilayer++) { | |
983 | for (itheta = 0; itheta < 180; itheta++) { | |
984 | sector = (TObjArray*)fNodes[ilayer][itheta]->At(isector); | |
985 | TObjArrayIter iter(sector); | |
986 | for (;;) { | |
987 | node = (AliITSnode*)iter.Next(); | |
988 | if (!node) break; | |
989 | node->InnerOf()->Clear(); | |
990 | node->OuterOf()->Clear(); | |
991 | /* | |
992 | delete node->InnerOf(); | |
993 | delete node->OuterOf(); | |
994 | node->InnerOf() = new TObjArray; | |
995 | node->OuterOf() = new TObjArray; | |
996 | */ | |
997 | } | |
998 | } | |
999 | } | |
1000 | } | |
1001 | ||
1002 | //__________________________________________________________________________________ | |
1003 | Int_t AliITStrackerANN::CreateNeurons | |
1004 | (AliITSnode *node, Int_t curvStep, Double_t vx, Double_t vy, Double_t vz) | |
1005 | { | |
1006 | // This method is used to create alle suitable neurons starting from | |
1007 | // a single AliITSnode. Each unit is also stored in the fInnerOf array | |
1008 | // of the passed node, and in the fOuterOf array of the other neuron edge. | |
1009 | // In the new implementation of the intermediate check steps, a further one | |
1010 | // is made, which chechs how well a helix is matched by three points | |
1011 | // in three consecutive layers. | |
1012 | // Then, a vertex-constrained check is made with vertex located | |
1013 | // in a layer L, for points in layers L+1 and L+2. | |
1014 | // | |
1015 | // In order to do this, the creator works recursively, in a tree-visit like operation. | |
1016 | // The neurons are effectively created only if the node argument passed is in | |
1017 | // the 5th layer (they are created between point of 5th and 6th layer). | |
1018 | // If the node is in an inner layer, its coordinates are passet as vertex for a nested | |
1019 | // call of the same function in the next two layers. | |
1020 | // | |
1021 | // Arguments: | |
1022 | // 1) reference node | |
1023 | // 2) current curvature cut step | |
1024 | // 3) X of referenced temporary (not primary) vertex | |
1025 | // 4) Y of referenced temporary (not primary) vertex | |
1026 | // 5) Z of referenced temporary (not primary) vertex | |
1027 | // | |
1028 | // Operations: | |
1029 | // - if the layer is the 5th, neurons are created with nodes | |
1030 | // in the fMatches array of the passed node | |
1031 | // - otherwise, the X, Y, Z of the passed node are given as | |
1032 | // vertex and the same procedure is recursively called for all | |
1033 | // nodes in the fMatches array of the passed one. | |
1034 | // | |
1035 | // Return values: | |
1036 | // - the total number of neurons created from the passed one | |
1037 | // summed with all neurons created from all nodes well matched with it | |
1038 | // (assumes a meaning only for nodes in the first layer) | |
1039 | ||
1040 | // local variables | |
1041 | Int_t made = 0; // counter | |
1042 | Bool_t found = 0; // flag | |
1043 | AliITSnode *match = 0; // cursor for a AliITSnode array | |
1044 | AliITSneuron *unit = 0; // cursor for a AliITSneuron array | |
1045 | ||
1046 | // --> Case 0: the passed node has already been used | |
1047 | // as member of a track found in a previous step. | |
1048 | // In this case, of course, the function exits. | |
1049 | if (node->IsUsed()) return 0; | |
1050 | ||
1051 | // --> Case 1: there are NO well-matched points. | |
1052 | // This can happen in all ITS layers, but it happens **for sure** | |
1053 | // for a node in the 'last' layer. | |
1054 | // Even in this case, the function exits. | |
1055 | if (node->Matches()->IsEmpty()) return 0; | |
1056 | ||
1057 | // --> Case 2: there are well-matched points. | |
1058 | // In this case, the function creates a neuron for each | |
1059 | // well-matched pair (according to the cuts for the current step) | |
1060 | // Moreover, before storing the neuron, a check is necessary | |
1061 | // to avoid the duplicate creation of the same neuron twice. | |
1062 | // (This could happen if the 3 last arguments of the function | |
1063 | // are close enough to cause a good match for the current step | |
1064 | // between two points, independently of their difference). | |
1065 | // Finally, a node is skipped if it has already been used. | |
1066 | // For each matched point for which a neuron is created, the procedure is | |
1067 | // recursively called. | |
1068 | TObjArrayIter matches(node->Matches()); | |
1069 | while ( (match = (AliITSnode*)matches.Next()) ) { | |
1070 | if (match->IsUsed()) continue; | |
1071 | if (!PassCurvCut(node, match, curvStep, vx, vy, vz)) continue; | |
1072 | found = kFALSE; | |
1073 | if (!node->InnerOf()->IsEmpty()) { | |
1074 | TObjArrayIter presentUnits(node->InnerOf()); | |
1075 | while ( (unit = (AliITSneuron*)presentUnits.Next()) ) { | |
1076 | if (unit->Inner() == node && unit->Outer() == match) { | |
1077 | found = kTRUE; | |
1078 | break; | |
1079 | } | |
1080 | } | |
1081 | } | |
1082 | if (found) continue; | |
1083 | AliITSneuron *unit = new AliITSneuron(node, match, fEdge2, fEdge1); | |
1084 | fNeurons->AddLast(unit); | |
1085 | node->InnerOf()->AddLast(unit); | |
1086 | match->OuterOf()->AddLast(unit); | |
1087 | made += CreateNeurons(match, curvStep, node->X(), node->Y(), node->Z()); | |
1088 | made++; | |
1089 | } | |
1090 | ||
1091 | // Of course, the return value contains the number of neurons | |
1092 | // counting in also the oned created in all levels of recursive calls. | |
1093 | return made; | |
1094 | } | |
1095 | ||
1096 | //__________________________________________________________________________________ | |
1097 | Int_t AliITStrackerANN::CreateNetwork(Int_t sector, Int_t curvStep) | |
1098 | { | |
1099 | // This function simply recalls the CreateNeurons() method for each node | |
1100 | // in the first layer, for the current sector. | |
1101 | // This generates the whole network, thanks to the recursive calls. | |
1102 | // | |
1103 | // Arguments: | |
1104 | // 1) current sector | |
1105 | // 2) current curvature step | |
1106 | // | |
1107 | // Operations: | |
1108 | // - scans the nodes array for all theta's in the current sector | |
1109 | // and layer 0, and calls the CreateNeurons() function. | |
1110 | ||
1111 | // removes all eventually present neurons | |
1112 | if (fNeurons) delete fNeurons; | |
1113 | fNeurons = new TObjArray; | |
1114 | fNeurons->SetOwner(kTRUE); | |
1115 | ||
1116 | // calls the ResetNodes() function to free the AliITSnode arrays | |
1117 | if (fMsgLevel >= 2) { | |
1118 | cout << "Sector " << sector << " PHI = "; | |
1119 | cout << fSectorWidth * (Double_t)sector << " --> "; | |
1120 | cout << fSectorWidth * (Double_t)(sector + 1) << endl; | |
1121 | cout << "Curvature step " << curvStep << " [cut = " << fCurvCut[curvStep] << "]" << endl; | |
1122 | } | |
1123 | ResetNodes(sector); | |
1124 | ||
1125 | // meaningful counters | |
1126 | Int_t itheta, neurons = 0; | |
1127 | TObjArray *lstSector = 0; | |
1128 | ||
1129 | // NEW VERSION | |
1130 | Double_t vx[6], vy[6], vz[6]; | |
1131 | AliITSnode *p[6] = {0, 0, 0, 0, 0, 0}; | |
1132 | for (itheta = 0; itheta < 180; itheta++) { | |
1133 | lstSector = (TObjArray*)fNodes[0][itheta]->At(sector); | |
1134 | TObjArrayIter lay0(lstSector); | |
1135 | while ( (p[0] = (AliITSnode*)lay0.Next()) ) { | |
1136 | if (p[0]->IsUsed()) continue; | |
1137 | vx[0] = fVertexX; | |
1138 | vy[0] = fVertexY; | |
1139 | vz[0] = fVertexZ; | |
1140 | neurons += CreateNeurons(p[0], curvStep, fVertexX, fVertexY, fVertexZ); | |
1141 | /* | |
1142 | TObjArrayIter lay1(p[0]->Matches()); | |
1143 | while ( (p[1] = (AliITSnode*)lay1.Next()) ) { | |
1144 | if (p[1]->IsUsed()) continue; | |
1145 | if (!PassCurvCut(p[0], p[1], curvStep, vx[0], vy[0], vz[0])) continue; | |
1146 | unit = new AliITSneuron; | |
1147 | unit->Inner() = p[0]; | |
1148 | unit->Outer() = p[1]; | |
1149 | unit->Activation() = gRandom->Rndm() * (fEdge1 - fEdge2) + fEdge2; | |
1150 | unit->fGain = new TObjArray; | |
1151 | fNeurons->AddLast(unit); | |
1152 | p[0]->InnerOf()->AddLast(unit); | |
1153 | p[1]->OuterOf()->AddLast(unit); | |
1154 | neurons++; | |
1155 | vx[1] = p[0]->X(); | |
1156 | vy[1] = p[0]->Y(); | |
1157 | vz[1] = p[0]->Z(); | |
1158 | TObjArrayIter lay2(p[1]->Matches()); | |
1159 | while ( (p[2] = (AliITSnode*)lay2.Next()) ) { | |
1160 | if (p[2]->IsUsed()) continue; | |
1161 | if (!PassCurvCut(p[1], p[2], curvStep, vx[1], vy[1], vz[1])) continue; | |
1162 | unit = new AliITSneuron; | |
1163 | unit->Inner() = p[1]; | |
1164 | unit->Outer() = p[2]; | |
1165 | unit->Activation() = gRandom->Rndm() * (fEdge1 - fEdge2) + fEdge2; | |
1166 | unit->fGain = new TObjArray; | |
1167 | fNeurons->AddLast(unit); | |
1168 | p[1]->InnerOf()->AddLast(unit); | |
1169 | p[2]->OuterOf()->AddLast(unit); | |
1170 | neurons++; | |
1171 | vx[2] = p[1]->X(); | |
1172 | vy[2] = p[1]->Y(); | |
1173 | vz[2] = p[1]->Z(); | |
1174 | TObjArrayIter lay3(p[2]->Matches()); | |
1175 | while ( (p[3] = (AliITSnode*)lay3.Next()) ) { | |
1176 | if (p[3]->IsUsed()) continue; | |
1177 | if (!PassCurvCut(p[2], p[3], curvStep, vx[2], vy[2], vz[2])) continue; | |
1178 | unit = new AliITSneuron; | |
1179 | unit->Inner() = p[2]; | |
1180 | unit->Outer() = p[3]; | |
1181 | unit->Activation() = gRandom->Rndm() * (fEdge1 - fEdge2) + fEdge2; | |
1182 | unit->fGain = new TObjArray; | |
1183 | fNeurons->AddLast(unit); | |
1184 | p[2]->InnerOf()->AddLast(unit); | |
1185 | p[3]->OuterOf()->AddLast(unit); | |
1186 | neurons++; | |
1187 | vx[3] = p[2]->X(); | |
1188 | vy[3] = p[2]->Y(); | |
1189 | vz[3] = p[2]->Z(); | |
1190 | TObjArrayIter lay4(p[3]->Matches()); | |
1191 | while ( (p[4] = (AliITSnode*)lay4.Next()) ) { | |
1192 | if (p[4]->IsUsed()) continue; | |
1193 | if (!PassCurvCut(p[3], p[4], curvStep, vx[3], vy[3], vz[3])) continue; | |
1194 | unit = new AliITSneuron; | |
1195 | unit->Inner() = p[3]; | |
1196 | unit->Outer() = p[4]; | |
1197 | unit->Activation() = gRandom->Rndm() * (fEdge1 - fEdge2) + fEdge2; | |
1198 | unit->fGain = new TObjArray; | |
1199 | fNeurons->AddLast(unit); | |
1200 | p[3]->InnerOf()->AddLast(unit); | |
1201 | p[4]->OuterOf()->AddLast(unit); | |
1202 | neurons++; | |
1203 | vx[4] = p[3]->X(); | |
1204 | vy[4] = p[3]->Y(); | |
1205 | vz[4] = p[3]->Z(); | |
1206 | TObjArrayIter lay5(p[4]->Matches()); | |
1207 | while ( (p[5] = (AliITSnode*)lay5.Next()) ) { | |
1208 | if (p[5]->IsUsed()) continue; | |
1209 | if (!PassCurvCut(p[4], p[5], curvStep, vx[4], vy[4], vz[4])) continue; | |
1210 | unit = new AliITSneuron; | |
1211 | unit->Inner() = p[4]; | |
1212 | unit->Outer() = p[5]; | |
1213 | unit->Activation() = gRandom->Rndm() * (fEdge1 - fEdge2) + fEdge2; | |
1214 | unit->fGain = new TObjArray; | |
1215 | fNeurons->AddLast(unit); | |
1216 | p[4]->InnerOf()->AddLast(unit); | |
1217 | p[5]->OuterOf()->AddLast(unit); | |
1218 | neurons++; | |
1219 | } // while (p[5]) | |
1220 | } // while (p[4]) | |
1221 | } // while (p[3]) | |
1222 | } // while (p[2]) | |
1223 | } // while (p[1]) | |
1224 | */ | |
1225 | } // while (p[0]) | |
1226 | } // for (itheta...) | |
1227 | // END OF NEW VERSION | |
1228 | ||
1229 | /* OLD VERSION | |
1230 | for (ilayer = 0; ilayer < 6; ilayer++) { | |
1231 | for (itheta = 0; itheta < 180; itheta++) { | |
1232 | lstSector = (TObjArray*)fNodes[ilayer][itheta]->At(sector_idx); | |
1233 | TObjArrayIter inners(lstSector); | |
1234 | while ( (inner = (AliITSnode*)inners.Next()) ) { | |
1235 | if (inner->GetUser() >= 0) continue; | |
1236 | TObjArrayIter outers(inner->Matches()); | |
1237 | while ( (outer = (AliITSnode*)outers.Next()) ) { | |
1238 | if (outer->GetUser() >= 0) continue; | |
1239 | if (!PassCurvCut(inner, outer, curvStep, fVX, fVY, fVZ)) continue; | |
1240 | unit = new AliITSneuron; | |
1241 | unit->Inner() = inner; | |
1242 | unit->Outer() = outer; | |
1243 | unit->Activation() = gRandom->Rndm() * (fEdge1 - fEdge2) + fEdge2; | |
1244 | unit->fGain = new TObjArray; | |
1245 | fNeurons->AddLast(unit); | |
1246 | inner->InnerOf()->AddLast(unit); | |
1247 | outer->OuterOf()->AddLast(unit); | |
1248 | neurons++; | |
1249 | } // for (;;) | |
1250 | } // for (;;) | |
1251 | } // for (itheta...) | |
1252 | } // for (ilayer...) | |
1253 | */ | |
1254 | ||
1255 | fNeurons->SetOwner(); | |
1256 | return neurons; | |
1257 | } | |
1258 | ||
1259 | //__________________________________________________________________________________ | |
1260 | Int_t AliITStrackerANN::LinkNeurons() | |
1261 | { | |
1262 | /*********************************************************************************** | |
1263 | ||
1264 | SYNAPSIS GENERATOR | |
1265 | ||
1266 | Scans the whole neuron array, in order to find all neuron pairs | |
1267 | which are connected in sequence and share a positive weight. | |
1268 | For each of them, an AliITSlink is created, which stores | |
1269 | the weight value, and will allow for a faster calculation | |
1270 | of the total neural input for each updating cycle. | |
1271 | ||
1272 | Every neuron contains an object array which stores all AliITSlink | |
1273 | objects which point to sequenced units, with the respective weights. | |
1274 | ||
1275 | Return value: | |
1276 | - the number of link created for the neural network. | |
1277 | If they are 0, no updating can be done and the step is skipped. | |
1278 | ||
1279 | ***********************************************************************************/ | |
1280 | ||
1281 | // meaningful indexes | |
1282 | Int_t total = 0; | |
1283 | Double_t weight = 0.0; | |
1284 | TObjArrayIter neurons(fNeurons), *iter; | |
1285 | AliITSneuron *neuron = 0, *test = 0; | |
1286 | ||
1287 | // scan in the neuron array | |
1288 | for (;;) { | |
1289 | neuron = (AliITSneuron*)neurons.Next(); | |
1290 | if (!neuron) break; | |
1291 | // checks for neurons startin from its head ( -> ) | |
1292 | iter = (TObjArrayIter*)neuron->Inner()->OuterOf()->MakeIterator(); | |
1293 | for (;;) { | |
1294 | test = (AliITSneuron*)iter->Next(); | |
1295 | if (!test) break; | |
1296 | weight = Weight(test, neuron); | |
1297 | if (weight > 0.0) neuron->Gain()->AddLast(new AliITSlink(weight, test)); | |
1298 | total++; | |
1299 | } | |
1300 | delete iter; | |
1301 | // checks for neurons ending in its tail ( >- ) | |
1302 | iter = (TObjArrayIter*)neuron->Outer()->InnerOf()->MakeIterator(); | |
1303 | for (;;) { | |
1304 | test = (AliITSneuron*)iter->Next(); | |
1305 | if (!test) break; | |
1306 | weight = Weight(neuron, test); | |
1307 | if (weight > 0.0) neuron->Gain()->AddLast(new AliITSlink(weight, test)); | |
1308 | total++; | |
1309 | } | |
1310 | delete iter; | |
1311 | } | |
1312 | return total; | |
1313 | } | |
1314 | ||
1315 | //__________________________________________________________________________________ | |
1316 | Bool_t AliITStrackerANN::Update() | |
1317 | { | |
1318 | /*********************************************************************************** | |
1319 | ||
1320 | Performs a single updating cycle. | |
1321 | ||
1322 | Operations: | |
1323 | - for each neuron, gets the activation with the neuron Activate() method | |
1324 | - checks if stability has been reached (compare mean activation variation | |
1325 | with the stability threshold data member) | |
1326 | ||
1327 | Return values: | |
1328 | - kTRUE means that the neural network has stabilized | |
1329 | - kFALSE means that another updating cycle is needed | |
1330 | ||
1331 | ***********************************************************************************/ | |
1332 | ||
1333 | Double_t actVar = 0.0, totDiff = 0.0; | |
1334 | TObjArrayIter iter(fNeurons); | |
1335 | AliITSneuron *unit; | |
1336 | for (;;) { | |
1337 | unit = (AliITSneuron*)iter.Next(); | |
1338 | if (!unit) break; | |
1339 | actVar = unit->Activate(fTemperature); | |
1340 | // calculation the relative activation variation | |
1341 | totDiff += actVar; | |
1342 | } | |
1343 | totDiff /= fNeurons->GetSize(); | |
1344 | return (totDiff < fStabThreshold); | |
1345 | } | |
1346 | ||
1347 | //__________________________________________________________________________________ | |
1348 | void AliITStrackerANN::FollowChains(Int_t sector) | |
1349 | { | |
1350 | /*********************************************************************************** | |
1351 | ||
1352 | CHAINS CREATION | |
1353 | ||
1354 | After that the neural network has stabilized, | |
1355 | the final step is to create polygonal chains | |
1356 | of clusters, one in each layer, which represent | |
1357 | the tracks recognized by the neural algorithm. | |
1358 | This is made by means of a choice of the best | |
1359 | neuron among the ones starting from each point. | |
1360 | ||
1361 | Once that such neuron is selected, its inner point | |
1362 | will set the 'fNext' field to its outer point, and | |
1363 | similarly, its outer point will set the 'fPrev' field | |
1364 | to its inner point. | |
1365 | This defines a bi-directional sequence. | |
1366 | ||
1367 | In this procedure, it can happen that many neurons | |
1368 | which have the head of the arrow in a given node, will | |
1369 | all select as best following the neuron with the largest | |
1370 | activation starting in that point. | |
1371 | This results in MANY nodes which have the same 'fNext'. | |
1372 | But, this field will be set to NULL for all these points, | |
1373 | but the only one which is pointed by the 'fPrev' field | |
1374 | of this shared node. | |
1375 | ||
1376 | ***********************************************************************************/ | |
1377 | ||
1378 | // meaningful counters | |
1379 | Int_t itheta, ilayer; | |
1380 | TObjArray *lstSector = 0; | |
1381 | Double_t test = fActMinimum; | |
1382 | AliITSnode *p = 0; | |
1383 | AliITSneuron *n = 0; | |
1384 | ||
1385 | // scan the whole collection of nodes | |
1386 | for (ilayer = 0; ilayer < fNLayers; ilayer++) { | |
1387 | for (itheta = 0; itheta < 180; itheta++) { | |
1388 | // get the array of point in a given layer/theta-slot/sector | |
1389 | lstSector = (TObjArray*)fNodes[ilayer][itheta]->At(sector); | |
1390 | TObjArrayIter nodes(lstSector); | |
1391 | while ( (p = (AliITSnode*)nodes.Next()) ) { | |
1392 | // if the point is used, it is skipped | |
1393 | if (p->IsUsed()) continue; | |
1394 | // initially, fNext points to nothing, and | |
1395 | // the comparison value is set to the minimum activation | |
1396 | // which allows to say that a neuron is turned 'on' | |
1397 | // a node from which only 'off' neurons start is probably | |
1398 | // a noise point, which will be excluded from the reconstruction. | |
1399 | test = fActMinimum; | |
1400 | p->Next() = 0; | |
1401 | TObjArrayIter innerof(p->InnerOf()); | |
1402 | while ( (n = (AliITSneuron*)innerof.Next()) ) { | |
1403 | // if the examined neuron has not the largest activation | |
1404 | // it is skipped and removed from array of all neurons | |
1405 | // and of its outer point (its inner is the cursor p) | |
1406 | if (n->Activation() < test) { | |
1407 | p->InnerOf()->Remove(n); | |
1408 | n->Outer()->OuterOf()->Remove(n); | |
1409 | delete fNeurons->Remove(n); | |
1410 | continue; | |
1411 | } | |
1412 | // otherwise, its activation becomes the maximum reference | |
1413 | p->Next() = n->Outer(); | |
1414 | // at the exit of the while(), the fNext will point | |
1415 | // to the outer node of the neuron starting in p, whose | |
1416 | // activation is the largest. | |
1417 | } | |
1418 | // the same procedure is made now for all neurons | |
1419 | // for which p is the outer point | |
1420 | test = fActMinimum; | |
1421 | p->Prev() = 0; | |
1422 | TObjArrayIter outerof(p->OuterOf()); | |
1423 | while ( (n = (AliITSneuron*)outerof.Next()) ) { | |
1424 | // if the examined neuron has not the largest activation | |
1425 | // it is skipped and removed from array of all neurons | |
1426 | // and of its inner point (its outer is the cursor p) | |
1427 | if (n->Activation() < test) { | |
1428 | p->OuterOf()->Remove(n); | |
1429 | n->Inner()->InnerOf()->Remove(n); | |
1430 | delete fNeurons->Remove(n); | |
1431 | continue; | |
1432 | } | |
1433 | // otherwise, its activation becomes the maximum reference | |
1434 | p->Prev() = n->Inner(); | |
1435 | // at the exit of the while(), the fPrev will point | |
1436 | // to the inner node of the neuron ending in p, whose | |
1437 | // activation is the largest. | |
1438 | } | |
1439 | } // end while (p ...) | |
1440 | } // end for (itheta ...) | |
1441 | } // end for (ilayer ...) | |
1442 | ||
1443 | // now the mismatches are solved | |
1444 | Bool_t matchPrev, matchNext; | |
1445 | for (ilayer = 0; ilayer < fNLayers; ilayer++) { | |
1446 | for (itheta = 0; itheta < 180; itheta++) { | |
1447 | // get the array of point in a given layer/theta-slot/sector | |
1448 | lstSector = (TObjArray*)fNodes[ilayer][itheta]->At(sector); | |
1449 | TObjArrayIter nodes(lstSector); | |
1450 | while ( (p = (AliITSnode*)nodes.Next()) ) { | |
1451 | // now p will point to a fPrev and a fNext node. | |
1452 | // Ideally they are placed this way: fPrev --> P --> fNext | |
1453 | // A mismatch happens if the point addressed as fPrev does NOT | |
1454 | // point to p as its fNext. And the same for the point addressed | |
1455 | // as fNext. | |
1456 | // In this case, the fNext and fPrev pointers are set to NULL | |
1457 | // and p is excluded from the reconstruction | |
1458 | matchPrev = matchNext= kFALSE; | |
1459 | if (ilayer > 0 && p->Prev() != NULL) | |
1460 | if (p->Prev()->Next() == p) matchPrev = kTRUE; | |
1461 | if (ilayer < 5 && p->Next() != NULL) | |
1462 | if (p->Next()->Prev() == p) matchNext = kTRUE; | |
1463 | if (ilayer == 0) | |
1464 | matchPrev = kTRUE; | |
1465 | else if (ilayer == 5) | |
1466 | matchNext = kTRUE; | |
1467 | if (!matchNext || !matchPrev) { | |
1468 | p->Prev() = p->Next() = 0; | |
1469 | } | |
1470 | } // end while (p ...) | |
1471 | } // end for (itheta ...) | |
1472 | } // end for (ilayer ...) | |
1473 | } | |
1474 | ||
1475 | //__________________________________________________________________________________ | |
1476 | Int_t AliITStrackerANN::SaveTracks(Int_t sector) | |
1477 | { | |
1478 | /******************************************************************************** | |
1479 | ||
1480 | TRACK SAVING | |
1481 | ------------ | |
1482 | Using the fNext and fPrev pointers, the chain is followed | |
1483 | and the track is fitted and saved. | |
1484 | Of course, the track is followed as a chain with a point | |
1485 | for each layer, then the track following starts always | |
1486 | from the clusters in layer 0. | |
1487 | ||
1488 | ***********************************************************************************/ | |
1489 | ||
1490 | // if not initialized, the tracks TobjArray is initialized | |
1491 | if (!fFoundTracks) fFoundTracks = new TObjArray; | |
1492 | ||
1493 | // meaningful counters | |
1494 | Int_t itheta, ilayer, l; | |
1495 | TObjArray *lstSector = 0; | |
1496 | AliITSnode *p = 0, *q = 0, **node = new AliITSnode*[fNLayers]; | |
1497 | for (l = 0; l < fNLayers; l++) node[l] = 0; | |
1498 | ||
1499 | /* | |
1500 | array = new AliITSnode*[fNLayers + 1]; | |
1501 | for (l = 0; l <= fNLayers; l++) array[l] = 0; | |
1502 | array[0] = new AliITSnode(); | |
1503 | array[0]->X() = fVertexX; | |
1504 | array[0]->Y() = fVertexY; | |
1505 | array[0]->Z() = fVertexZ; | |
1506 | array[0]->ErrX2() = fVertexErrorX; | |
1507 | array[0]->ErrY2() = fVertexErrorY; | |
1508 | array[0]->ErrZ2() = fVertexErrorZ; | |
1509 | */ | |
1510 | Double_t *param = new Double_t[8]; | |
1511 | ||
1512 | // scan the whole collection of nodes | |
1513 | for (ilayer = 0; ilayer < 1; ilayer++) { | |
1514 | for (itheta = 0; itheta < 180; itheta++) { | |
1515 | // get the array of point in a given layer/theta-slot/sector | |
1516 | lstSector = (TObjArray*)fNodes[ilayer][itheta]->At(sector); | |
1517 | TObjArrayIter nodes(lstSector); | |
1518 | while ( (p = (AliITSnode*)nodes.Next()) ) { | |
1519 | for (q = p; q; q = q->Next()) { | |
1520 | l = q->GetLayer(); | |
1521 | node[l] = q; | |
1522 | } | |
1523 | //if (!RiemannFit(fNLayers, node, param)) continue; | |
1524 | // initialization of Kalman Filter Tracking | |
00a7cc50 | 1525 | AliITSRecPoint *cluster = (AliITSRecPoint*)GetCluster(node[0]->ClusterRef()); |
cec46807 | 1526 | Int_t mod = cluster->GetDetectorIndex(); |
1527 | Int_t lay, lad, det; | |
1528 | fGeom->GetModuleId(mod, lay, lad, det); | |
1529 | Float_t y0 = cluster->GetY(); | |
1530 | Float_t z0 = cluster->GetZ(); | |
7d62fb64 | 1531 | AliITStrackSA* trac = new AliITStrackSA(fGeom,lay, lad, det, |
cec46807 | 1532 | y0, z0, |
1533 | param[4], param[7], param[3], 1); | |
1534 | for (l = 0; l < fNLayers; l++) { | |
00a7cc50 | 1535 | cluster = (AliITSRecPoint*)GetCluster(node[l]->ClusterRef()); |
cec46807 | 1536 | if (cluster) trac->AddClusterV2(l, (node[l]->ClusterRef() & 0x0fffffff)>>0); |
1537 | } | |
1538 | AliITStrackV2* ot = new AliITStrackV2(*trac); | |
6c94f330 | 1539 | ot->ResetCovariance(10.); |
cec46807 | 1540 | ot->ResetClusters(); |
1541 | if (RefitAt(49.,ot,trac)) { //fit from layer 1 to layer 6 | |
1542 | AliITStrackV2 *otrack2 = new AliITStrackV2(*ot); | |
6c94f330 | 1543 | otrack2->ResetCovariance(10.); |
cec46807 | 1544 | otrack2->ResetClusters(); |
1545 | //fit from layer 6 to layer 1 | |
1546 | if (RefitAt(3.7,otrack2,ot)) fFoundTracks->AddLast(otrack2); | |
1547 | } | |
1548 | // end of Kalman Filter fit | |
1549 | } | |
1550 | } | |
1551 | } | |
1552 | ||
1553 | return 1; | |
1554 | } | |
1555 | ||
1556 | //__________________________________________________________________________________ | |
1557 | void AliITStrackerANN::ExportTracks(const char *filename) const | |
1558 | { | |
1559 | // Exports found tracks into a TTree of AliITStrackV2 objects | |
1560 | TFile *file = new TFile(filename, "RECREATE"); | |
1561 | TTree *tree = new TTree("TreeT-ANN", "Tracks found in ITS stand-alone with Neural Tracking"); | |
1562 | AliITStrackV2 *track = 0; | |
1563 | tree->Branch("Tracks", &track, "AliITStrackV2"); | |
1564 | TObjArrayIter tracks(fFoundTracks); | |
1565 | while ( (track = (AliITStrackV2*)tracks.Next()) ) { | |
1566 | tree->Fill(); | |
1567 | } | |
1568 | file->cd(); | |
1569 | tree->Write(); | |
1570 | file->Close(); | |
1571 | } | |
1572 | ||
1573 | ||
1574 | //__________________________________________________________________________________ | |
1575 | void AliITStrackerANN::CleanNetwork() | |
1576 | { | |
1577 | // Removes deactivated units from the network | |
1578 | ||
1579 | AliITSneuron *unit = 0; | |
1580 | TObjArrayIter neurons(fNeurons); | |
1581 | while ( (unit = (AliITSneuron*)neurons.Next()) ) { | |
1582 | if (unit->Activation() < fActMinimum) { | |
1583 | unit->Inner()->InnerOf()->Remove(unit); | |
1584 | unit->Outer()->OuterOf()->Remove(unit); | |
1585 | delete fNeurons->Remove(unit); | |
1586 | } | |
1587 | } | |
1588 | return; | |
1589 | Bool_t removed; | |
1590 | Int_t nIn, nOut; | |
1591 | AliITSneuron *enemy = 0; | |
1592 | neurons.Reset(); | |
1593 | while ( (unit = (AliITSneuron*)neurons.Next()) ) { | |
1594 | nIn = (Int_t)unit->Inner()->InnerOf()->GetSize(); | |
1595 | nOut = (Int_t)unit->Outer()->OuterOf()->GetSize(); | |
1596 | if (nIn < 2 && nOut < 2) continue; | |
1597 | removed = kFALSE; | |
1598 | if (nIn > 1) { | |
1599 | TObjArrayIter competing(unit->Inner()->InnerOf()); | |
1600 | while ( (enemy = (AliITSneuron*)competing.Next()) ) { | |
1601 | if (unit->Activation() > enemy->Activation()) { | |
1602 | enemy->Inner()->InnerOf()->Remove(enemy); | |
1603 | enemy->Outer()->OuterOf()->Remove(enemy); | |
1604 | delete fNeurons->Remove(enemy); | |
1605 | } | |
1606 | else { | |
1607 | unit->Inner()->InnerOf()->Remove(unit); | |
1608 | unit->Outer()->OuterOf()->Remove(unit); | |
1609 | delete fNeurons->Remove(unit); | |
1610 | removed = kTRUE; | |
1611 | break; | |
1612 | } | |
1613 | } | |
1614 | if (removed) continue; | |
1615 | } | |
1616 | if (nOut > 1) { | |
1617 | TObjArrayIter competing(unit->Outer()->OuterOf()); | |
1618 | while ( (enemy = (AliITSneuron*)competing.Next()) ) { | |
1619 | if (unit->Activation() > enemy->Activation()) { | |
1620 | enemy->Inner()->InnerOf()->Remove(enemy); | |
1621 | enemy->Outer()->OuterOf()->Remove(enemy); | |
1622 | delete fNeurons->Remove(enemy); | |
1623 | } | |
1624 | else { | |
1625 | unit->Inner()->InnerOf()->Remove(unit); | |
1626 | unit->Outer()->OuterOf()->Remove(unit); | |
1627 | delete fNeurons->Remove(unit); | |
1628 | removed = kTRUE; | |
1629 | break; | |
1630 | } | |
1631 | } | |
1632 | } | |
1633 | } | |
1634 | } | |
1635 | ||
1636 | //__________________________________________________________________________________ | |
1637 | Int_t AliITStrackerANN::StoreTracks() | |
1638 | { | |
1639 | // Stores the tracks found in a single neural tracking step. | |
1640 | // In order to do this, it sects each neuron which has a point | |
1641 | // in the first layer. | |
1642 | // Then | |
1643 | ||
1644 | // if not initialized, the tracks TobjArray is initialized | |
1645 | if (!fFoundTracks) fFoundTracks = new TObjArray; | |
1646 | ||
1647 | Int_t i, check, stored = 0; | |
1648 | Double_t testAct = 0; | |
1649 | AliITSneuron *unit = 0, *cursor = 0, *fwd = 0; | |
1650 | AliITSnode *node = 0; | |
1651 | TObjArrayIter iter(fNeurons), *fwdIter; | |
1652 | TObjArray *removedUnits = new TObjArray(0); | |
1653 | removedUnits->SetOwner(kFALSE); | |
1654 | AliITStrackANN annTrack(fNLayers); | |
1655 | ||
1656 | for (;;) { | |
1657 | unit = (AliITSneuron*)iter.Next(); | |
1658 | if (!unit) break; | |
1659 | if (unit->Inner()->GetLayer() > 0) continue; | |
1660 | annTrack.SetNode(unit->Inner()->GetLayer(), unit->Inner()); | |
1661 | annTrack.SetNode(unit->Outer()->GetLayer(), unit->Outer()); | |
1662 | node = unit->Outer(); | |
1663 | removedUnits->AddLast(unit); | |
1664 | while (node) { | |
1665 | testAct = fActMinimum; | |
1666 | fwdIter = (TObjArrayIter*)node->InnerOf()->MakeIterator(); | |
1667 | fwd = 0; | |
1668 | for (;;) { | |
1669 | cursor = (AliITSneuron*)fwdIter->Next(); | |
1670 | if (!cursor) break; | |
1671 | if (cursor->Used()) continue; | |
1672 | if (cursor->Activation() >= testAct) { | |
1673 | testAct = cursor->Activation(); | |
1674 | fwd = cursor; | |
1675 | } | |
1676 | } | |
1677 | if (!fwd) break; | |
1678 | removedUnits->AddLast(fwd); | |
1679 | node = fwd->Outer(); | |
1680 | annTrack.SetNode(node->GetLayer(), node); | |
1681 | } | |
1682 | check = annTrack.CheckOccupation(); | |
1683 | if (check >= 6) { | |
1684 | stored++; | |
1685 | // FIT | |
1686 | //if (!RiemannFit(fNLayers, trackitem, param)) continue; | |
1687 | if (!annTrack.RiemannFit()) continue; | |
1688 | // initialization of Kalman Filter Tracking | |
00a7cc50 | 1689 | AliITSRecPoint *cluster = (AliITSRecPoint*)GetCluster(annTrack[0]->ClusterRef()); |
cec46807 | 1690 | Int_t mod = cluster->GetDetectorIndex(); |
1691 | Int_t lay, lad, det; | |
1692 | fGeom->GetModuleId(mod, lay, lad, det); | |
1693 | Float_t y0 = cluster->GetY(); | |
1694 | Float_t z0 = cluster->GetZ(); | |
7d62fb64 | 1695 | AliITStrackSA* trac = new AliITStrackSA(fGeom,lay, lad, det, y0, z0, |
cec46807 | 1696 | annTrack.Phi(), annTrack.TanLambda(), |
1697 | annTrack.Curv(), 1); | |
1698 | for (Int_t l = 0; l < fNLayers; l++) { | |
1699 | if (!annTrack[l]) continue; | |
00a7cc50 | 1700 | cluster = (AliITSRecPoint*)GetCluster(annTrack[l]->ClusterRef()); |
cec46807 | 1701 | if (cluster) trac->AddClusterV2(l, (annTrack[l]->ClusterRef() & 0x0fffffff)>>0); |
1702 | } | |
1703 | AliITStrackV2* ot = new AliITStrackV2(*trac); | |
6c94f330 | 1704 | ot->ResetCovariance(10.); |
cec46807 | 1705 | ot->ResetClusters(); |
1706 | if (RefitAt(49.,ot,trac)) { //fit from layer 1 to layer 6 | |
1707 | AliITStrackV2 *otrack2 = new AliITStrackV2(*ot); | |
6c94f330 | 1708 | otrack2->ResetCovariance(10.); |
cec46807 | 1709 | otrack2->ResetClusters(); |
1710 | //fit from layer 6 to layer 1 | |
1711 | if (RefitAt(3.7,otrack2,ot)) fFoundTracks->AddLast(otrack2); | |
1712 | } | |
1713 | // end of Kalman Filter fit | |
1714 | // END FIT | |
1715 | for (i = 0; i < fNLayers; i++) { | |
1716 | //node = (AliITSnode*)removedPoints->At(i); | |
1717 | //node->Use(); | |
1718 | annTrack[i]->Use(); | |
1719 | } | |
1720 | fwdIter = (TObjArrayIter*)removedUnits->MakeIterator(); | |
1721 | for (;;) { | |
1722 | cursor = (AliITSneuron*)fwdIter->Next(); | |
1723 | if(!cursor) break; | |
1724 | cursor->Used() = 1; | |
1725 | } | |
1726 | } | |
1727 | } | |
1728 | ||
1729 | return stored; | |
1730 | } | |
1731 | ||
1732 | Double_t AliITStrackerANN::Weight(AliITSneuron *nAB, AliITSneuron *nBC) | |
1733 | { | |
1734 | /*********************************************************************************** | |
1735 | * Calculation of neural weight. | |
1736 | * The implementation of positive neural weight is set only in the case | |
1737 | * of connected units (e.g.: A->B with B->C). | |
1738 | * Given that B is the **common** point. care should be taken to pass | |
1739 | * as FIRST argument the neuron going "to" B, and | |
1740 | * as SECOND argument the neuron starting "from" B | |
1741 | * anyway, a check is put in order to return 0.0 when arguments are not well given. | |
1742 | ***********************************************************************************/ | |
1743 | ||
1744 | if (nAB->Outer() != nBC->Inner()) { | |
1745 | if (nBC->Outer() == nAB->Inner()) { | |
1746 | AliITSneuron *temp = nAB; | |
1747 | nAB = nBC; | |
1748 | nBC = temp; | |
1749 | temp = 0; | |
1750 | if (fMsgLevel >= 3) { | |
1751 | Info("Weight", "Switching wrongly ordered arguments."); | |
1752 | } | |
1753 | } | |
1754 | Warning("Weight", "Not connected segments. Returning 0.0"); | |
1755 | return 0.0; | |
1756 | } | |
1757 | ||
1758 | AliITSnode *pA = nAB->Inner(); | |
1759 | AliITSnode *pB = nAB->Outer(); | |
1760 | AliITSnode *pC = nBC->Outer(); | |
1761 | ||
1762 | TVector3 vAB(pB->X() - pA->X(), pB->Y() - pA->Y(), pB->Z() - pA->Z()); | |
1763 | TVector3 vBC(pC->X() - pB->X(), pC->Y() - pB->Y(), pC->Z() - pB->Z()); | |
1764 | ||
1765 | Double_t weight = 1.0 - sin(vAB.Angle(vBC)); | |
1766 | return fGain2CostRatio * TMath::Power(weight, fExponent); | |
1767 | } | |
1768 | ||
1769 | ||
1770 | ||
1771 | /****************************************** | |
1772 | ****************************************** | |
1773 | *** AliITStrackerANN::AliITSnode class *** | |
1774 | ****************************************** | |
1775 | ******************************************/ | |
1776 | ||
1777 | //__________________________________________________________________________________ | |
cb729508 | 1778 | AliITStrackerANN::AliITSnode::AliITSnode(): |
1779 | fX(0.0), | |
1780 | fY(0.0), | |
1781 | fZ(0.0), | |
1782 | fEX2(0.0), | |
1783 | fEY2(0.0), | |
1784 | fEZ2(0.0), | |
1785 | fUsed(kFALSE), | |
1786 | fClusterRef(-1), | |
1787 | fMatches(NULL), | |
1788 | fInnerOf(NULL), | |
1789 | fOuterOf(NULL), | |
1790 | fNext(NULL), | |
1791 | fPrev(NULL) | |
cec46807 | 1792 | { |
1793 | // Constructor for the embedded 'AliITSnode' class. | |
1794 | // It initializes all pointer-like objects. | |
1795 | ||
cec46807 | 1796 | } |
1797 | ||
cb729508 | 1798 | AliITStrackerANN::AliITSnode::AliITSnode(const AliITSnode &n) : TObject((TObject&)n), |
1799 | fX(n.fX), | |
1800 | fY(n.fY), | |
1801 | fZ(n.fZ), | |
1802 | fEX2(n.fEX2), | |
1803 | fEY2(n.fEY2), | |
1804 | fEZ2(n.fEZ2), | |
1805 | fUsed(n.fUsed), | |
1806 | fClusterRef(n.fClusterRef), | |
1807 | fMatches(n.fMatches), | |
1808 | fInnerOf(n.fInnerOf), | |
1809 | fOuterOf(n.fOuterOf), | |
1810 | fNext(n.fNext), | |
1811 | fPrev(n.fPrev){ | |
1812 | //copy constructor | |
1813 | } | |
cec46807 | 1814 | //__________________________________________________________________________________ |
1815 | AliITStrackerANN::AliITSnode::~AliITSnode() | |
1816 | { | |
1817 | // Destructor for the embedded 'AliITSnode' class. | |
1818 | // It should clear the object arrays, but it is possible | |
1819 | // that some objects still are useful after the point deletion | |
1820 | // then the arrays are cleared but their objects are owed by | |
1821 | // another TCollection object, and not deleted. | |
1822 | // For safety reasons, all the pointers are set to zero. | |
1823 | ||
1824 | fInnerOf->SetOwner(kFALSE); | |
1825 | fInnerOf->Clear(); | |
1826 | delete fInnerOf; | |
1827 | fInnerOf = 0; | |
1828 | fOuterOf->SetOwner(kFALSE); | |
1829 | fOuterOf->Clear(); | |
1830 | delete fOuterOf; | |
1831 | fOuterOf = 0; | |
1832 | fMatches->SetOwner(kFALSE); | |
1833 | fMatches->Clear(); | |
1834 | delete fMatches; | |
1835 | fMatches = 0; | |
1836 | fNext = 0; | |
1837 | fPrev = 0; | |
1838 | } | |
1839 | ||
1840 | //__________________________________________________________________________________ | |
0db9364f | 1841 | Double_t AliITStrackerANN::AliITSnode::GetPhi() const |
cec46807 | 1842 | { |
1843 | // Calculates the 'phi' (azimutal) angle, and returns it | |
1844 | // in the range between 0 and 2Pi radians. | |
1845 | ||
1846 | Double_t q; | |
1847 | q = TMath::ATan2(fY,fX); | |
1848 | if (q >= 0.) | |
1849 | return q; | |
1850 | else | |
1851 | return q + TMath::TwoPi(); | |
1852 | } | |
1853 | ||
1854 | //__________________________________________________________________________________ | |
0db9364f | 1855 | Double_t AliITStrackerANN::AliITSnode::GetError(Option_t *option) |
cec46807 | 1856 | { |
1857 | // Returns the error or the square error of | |
1858 | // values related to the coordinates in different systems. | |
1859 | // The option argument specifies the coordinate error desired: | |
1860 | // | |
1861 | // "R2" --> error in transverse radius | |
1862 | // "R3" --> error in spherical radius | |
1863 | // "PHI" --> error in azimuthal angle | |
1864 | // "THETA" --> error in polar angle | |
1865 | // "SQ" --> get the square of error | |
1866 | // | |
1867 | // In order to get the error on the cartesian coordinates | |
1868 | // reference to the inline ErrX2(), ErrY2() adn ErrZ2() methods. | |
1869 | ||
1870 | TString opt(option); | |
1871 | Double_t errorSq = 0.0; | |
1872 | opt.ToUpper(); | |
1873 | ||
1874 | if (opt.Contains("R2")) { | |
1875 | errorSq = fX*fX*fEX2 + fY*fY*fEY2; | |
1876 | errorSq /= GetR2sq(); | |
1877 | } | |
1878 | else if (opt.Contains("R3")) { | |
1879 | errorSq = fX*fX*fEX2 + fY*fY*fEY2 + fZ*fZ*fEZ2; | |
1880 | errorSq /= GetR3sq(); | |
1881 | } | |
1882 | else if (opt.Contains("PHI")) { | |
1883 | errorSq = fY*fY*fEX2; | |
1884 | errorSq += fX*fX*fEY2; | |
1885 | errorSq /= GetR2sq() * GetR2sq(); | |
1886 | } | |
1887 | else if (opt.Contains("THETA")) { | |
1888 | errorSq = fZ*fZ * (fX*fX*fEX2 + fY*fY*fEY2); | |
1889 | errorSq += GetR2sq() * GetR2sq() * fEZ2; | |
1890 | errorSq /= GetR3sq() * GetR3sq() * GetR2() * GetR2(); | |
1891 | } | |
1892 | ||
1893 | if (!opt.Contains("SQ")) | |
1894 | return TMath::Sqrt(errorSq); | |
1895 | else | |
1896 | return errorSq; | |
1897 | } | |
1898 | ||
1899 | ||
1900 | ||
1901 | /******************************************** | |
1902 | ******************************************** | |
1903 | *** AliITStrackerANN::AliITSneuron class *** | |
1904 | ******************************************** | |
1905 | ********************************************/ | |
1906 | ||
1907 | //__________________________________________________________________________________ | |
1908 | AliITStrackerANN::AliITSneuron::AliITSneuron | |
cb729508 | 1909 | (AliITSnode *inner, AliITSnode *outer, Double_t minAct, Double_t maxAct): |
1910 | fUsed(0), | |
1911 | fActivation(0), | |
1912 | fInner(inner), | |
1913 | fOuter(outer), | |
1914 | fGain(0) | |
cec46807 | 1915 | { |
1916 | // Default neuron constructor | |
1917 | fActivation = gRandom->Rndm() * (maxAct-minAct) + minAct; | |
1918 | fGain = new TObjArray; | |
1919 | } | |
cb729508 | 1920 | |
1921 | AliITStrackerANN::AliITSneuron::AliITSneuron(const AliITSneuron &n) : TObject((TObject&)n), | |
1922 | fUsed(n.fUsed), | |
1923 | fActivation(n.fActivation), | |
1924 | fInner(n.fInner), | |
1925 | fOuter(n.fOuter), | |
1926 | fGain(n.fGain){ | |
1927 | //copy constructor | |
1928 | } | |
cec46807 | 1929 | //__________________________________________________________________________________ |
0db9364f | 1930 | Double_t AliITStrackerANN::AliITSneuron::Activate(Double_t temperature) |
cec46807 | 1931 | { |
1932 | // This computes the new activation of a neuron, and returns | |
1933 | // its activation variation as a consequence of the updating. | |
1934 | // | |
1935 | // Arguments: | |
1936 | // - the 'temperature' parameter for the neural activation logistic function | |
1937 | // | |
1938 | // Operations: | |
1939 | // - collects the total gain, by summing the products | |
1940 | // of the activation of each sequenced unit by the relative weight. | |
1941 | // - collects the total cost, by summing the activations of | |
1942 | // all competing units | |
1943 | // - passes the sum of gain - cost to the activation function and | |
1944 | // calculates the new activation | |
1945 | // | |
1946 | // Return value: | |
1947 | // - the difference between the old activation and the new one | |
1948 | // (absolute value) | |
1949 | ||
1950 | // local variables | |
1951 | Double_t sumGain = 0.0; // total contribution from chained neurons | |
1952 | Double_t sumCost = 0.0; // total contribution from crossing neurons | |
1953 | Double_t input; // total input | |
1954 | Double_t actOld, actNew; // old and new values for the activation | |
1955 | AliITSneuron *linked = 0; // cursor for scanning the neuron arrays (for link check) | |
1956 | AliITSlink *link; // cursor for scanning the synapses arrays (for link check) | |
1957 | TObjArrayIter *iterator = 0; // pointer to the iterator along the neuron arrays | |
1958 | ||
1959 | // sum contributions from the correlated units | |
1960 | iterator = (TObjArrayIter*)fGain->MakeIterator(); | |
1961 | for(;;) { | |
1962 | link = (AliITSlink*)iterator->Next(); | |
1963 | if (!link) break; | |
1964 | sumGain += link->Contribution(); | |
1965 | } | |
1966 | delete iterator; | |
1967 | ||
1968 | // sum contributions from the competing units: | |
1969 | // the ones which have the same starting point... | |
1970 | iterator = (TObjArrayIter*)fInner->InnerOf()->MakeIterator(); | |
1971 | for (;;) { | |
1972 | linked = (AliITSneuron*)iterator->Next(); | |
1973 | if (!linked) break; | |
1974 | if (linked == this) continue; | |
1975 | sumCost += linked->fActivation; | |
1976 | } | |
1977 | delete iterator; | |
1978 | // ...and the ones which have the same ending point | |
1979 | iterator = (TObjArrayIter*)fOuter->OuterOf()->MakeIterator(); | |
1980 | for (;;) { | |
1981 | linked = (AliITSneuron*)iterator->Next(); | |
1982 | if (!linked) break; | |
1983 | if (linked == this) continue; | |
1984 | sumCost += linked->fActivation; | |
1985 | } | |
1986 | ||
1987 | // calculate the total input as the difference between gain and cost | |
1988 | input = (sumGain - sumCost) / temperature; | |
1989 | actOld = fActivation; | |
1990 | // calculate the final output | |
1991 | #ifdef NEURAL_LINEAR | |
1992 | if (input <= -2.0 * temperature) | |
1993 | actNew = 0.0; | |
1994 | else if (input >= 2.0 * temperature) | |
1995 | actNew = 1.0; | |
1996 | else | |
1997 | actNew = input / (4.0 * temperature) + 0.5; | |
1998 | #else | |
1999 | actNew = 1.0 / (1.0 + TMath::Exp(-input)); | |
2000 | #endif | |
2001 | fActivation = actNew; | |
2002 | ||
2003 | // return the activation variation | |
2004 | return TMath::Abs(actNew - actOld); | |
2005 | } | |
2006 | ||
2007 | ||
2008 | ||
2009 | /****************************************** | |
2010 | ****************************************** | |
2011 | *** AliITStrackerANN::AliITSlink class *** | |
2012 | ****************************************** | |
2013 | ******************************************/ | |
2014 | ||
2015 | // No methods defined non-inline | |
2016 | ||
2017 | ||
2018 | ||
2019 | /********************************************** | |
2020 | ********************************************** | |
2021 | *** AliITStrackerANN::AliITStrackANN class *** | |
2022 | ********************************************** | |
2023 | **********************************************/ | |
2024 | ||
2025 | //__________________________________________________________________________________ | |
cb729508 | 2026 | AliITStrackerANN::AliITStrackANN::AliITStrackANN(Int_t dim) : |
2027 | fNPoints(dim), | |
2028 | fXCenter(0.0), | |
2029 | fYCenter(0.0), | |
2030 | fRadius(0.0), | |
2031 | fCurv(0.0), | |
2032 | fDTrans(0.0), | |
2033 | fDLong(0.0), | |
2034 | fTanLambda(0.0), | |
2035 | fPhi(0.0), | |
2036 | fNode(0) | |
cec46807 | 2037 | { |
2038 | // Default constructor for the AliITStrackANN class | |
2039 | ||
cec46807 | 2040 | if (! dim) { |
2041 | fNode = 0; | |
2042 | } | |
2043 | else{ | |
2044 | Int_t i = 0; | |
2045 | fNode = new AliITSnode*[dim]; | |
2046 | for (i = 0; i < dim; i++) fNode[i] = 0; | |
2047 | } | |
2048 | } | |
2049 | ||
cb729508 | 2050 | AliITStrackerANN::AliITStrackANN::AliITStrackANN(const AliITStrackANN &n) : TObject((TObject&)n), |
2051 | fNPoints(n.fNPoints), | |
2052 | fXCenter(n.fXCenter), | |
2053 | fYCenter(n.fYCenter), | |
2054 | fRadius(n.fRadius), | |
2055 | fCurv(n.fCurv), | |
2056 | fDTrans(n.fDTrans), | |
2057 | fDLong(n.fDLong), | |
2058 | fTanLambda(n.fTanLambda), | |
2059 | fPhi(n.fPhi), | |
2060 | fNode(n.fNode) | |
2061 | { | |
2062 | //copy constructor | |
2063 | } | |
cec46807 | 2064 | //__________________________________________________________________________________ |
2065 | Int_t AliITStrackerANN::AliITStrackANN::CheckOccupation() const | |
2066 | { | |
2067 | // Returns the number of pointers fNode which are not NULL | |
2068 | ||
2069 | Int_t i; // cursor | |
2070 | Int_t count = 0; // counter for how many points are stored in the track | |
2071 | ||
2072 | for (i = 0; i < fNPoints; i++) { | |
2073 | if (fNode[i] != NULL) count++; | |
2074 | } | |
2075 | ||
2076 | return count; | |
2077 | } | |
2078 | ||
2079 | //__________________________________________________________________________________ | |
2080 | Bool_t AliITStrackerANN::AliITStrackANN::RiemannFit() | |
2081 | { | |
2082 | // Performs the Riemann Sphere fit for the given points to a circle | |
2083 | // and then uses the fit parameters to fit a helix in space. | |
2084 | // | |
2085 | // Return values: | |
2086 | // - kTRUE if all operations have been performed | |
2087 | // - kFALSE if the numbers risk to lead to an arithmetic violation | |
2088 | ||
2089 | Int_t i, j, count, dim = fNPoints; | |
2090 | ||
2091 | // First check for all points | |
2092 | count = CheckOccupation(); | |
2093 | if (count != fNPoints) { | |
2094 | Error ("AliITStrackANN::RiemannFit", "CheckOccupations returns %d, fNPoints = %d ==> MISMATCH", count, fNPoints); | |
2095 | return kFALSE; | |
2096 | } | |
2097 | ||
2098 | // matrix of ones | |
2099 | TMatrixD m1(dim,1); | |
2100 | for (i = 0; i < dim; i++) m1(i,0) = 1.0; | |
2101 | ||
2102 | // matrix of Rieman projection coordinates | |
2103 | TMatrixD coords(dim,3); | |
2104 | for (i = 0; i < dim; i++) { | |
2105 | coords(i,0) = fNode[i]->X(); | |
2106 | coords(i,1) = fNode[i]->Y(); | |
2107 | coords(i,2) = fNode[i]->GetR2sq(); | |
2108 | } | |
2109 | ||
2110 | // matrix of weights | |
2111 | Double_t xterm, yterm, ex, ey; | |
2112 | TMatrixD weights(dim,dim); | |
2113 | for (i = 0; i < dim; i++) { | |
2114 | xterm = fNode[i]->X() * fNode[i]->GetPhi() - fNode[i]->Y() / fNode[i]->GetR2(); | |
2115 | ex = fNode[i]->ErrX2(); | |
2116 | yterm = fNode[i]->Y() * fNode[i]->GetPhi() + fNode[i]->X() / fNode[i]->GetR2(); | |
2117 | ey = fNode[i]->ErrY2(); | |
2118 | weights(i,i) = fNode[i]->GetR2sq() / (xterm * xterm * ex + yterm * yterm * ey ); | |
2119 | } | |
2120 | ||
2121 | // weighted sample mean | |
2122 | Double_t meanX = 0.0, meanY = 0.0, meanW = 0.0, sw = 0.0; | |
2123 | for (i = 0; i < dim; i++) { | |
2124 | meanX += weights(i,i) * coords(i,0); | |
2125 | meanY += weights(i,i) * coords(i,1); | |
2126 | meanW += weights(i,i) * coords(i,2); | |
2127 | sw += weights(i,i); | |
2128 | } | |
2129 | meanX /= sw; | |
2130 | meanY /= sw; | |
2131 | meanW /= sw; | |
2132 | ||
2133 | // sample covariance matrix | |
2134 | for (i = 0; i < dim; i++) { | |
2135 | coords(i,0) -= meanX; | |
2136 | coords(i,1) -= meanY; | |
2137 | coords(i,2) -= meanW; | |
2138 | } | |
2139 | TMatrixD coordsT(TMatrixD::kTransposed, coords); | |
2140 | TMatrixD weights4coords(weights, TMatrixD::kMult, coords); | |
2141 | TMatrixD sampleCov(coordsT, TMatrixD::kMult, weights4coords); | |
2142 | for (i = 0; i < 3; i++) { | |
2143 | for (j = i + 1; j < 3; j++) { | |
2144 | sampleCov(i,j) = sampleCov(j,i) = (sampleCov(i,j) + sampleCov(j,i)) * 0.5; | |
2145 | } | |
2146 | } | |
2147 | ||
2148 | // Eigenvalue problem solving for V matrix | |
2149 | Int_t ileast = 0; | |
2150 | TVectorD eval(3), n(3); | |
292a2409 | 2151 | #if ROOT_VERSION_CODE < ROOT_VERSION(4,0,2) |
cec46807 | 2152 | TMatrixD evec = sampleCov.EigenVectors(eval); |
292a2409 | 2153 | #else |
2154 | TMatrixDEigen ei(sampleCov); | |
2155 | TMatrixD evec = ei.GetEigenVectors(); | |
2156 | eval = ei.GetEigenValuesRe(); | |
2157 | #endif | |
cec46807 | 2158 | if (eval(1) < eval(ileast)) ileast = 1; |
2159 | if (eval(2) < eval(ileast)) ileast = 2; | |
2160 | n(0) = evec(0, ileast); | |
2161 | n(1) = evec(1, ileast); | |
2162 | n(2) = evec(2, ileast); | |
2163 | ||
2164 | // c - known term in the plane intersection with Riemann axes | |
2165 | Double_t c = -(meanX * n(0) + meanY * n(1) + meanW * n(2)); | |
2166 | ||
2167 | // center and radius of fitted circle | |
2168 | Double_t xc, yc, radius, curv; | |
2169 | xc = -n(0) / (2. * n(2)); | |
2170 | yc = -n(1) / (2. * n(2)); | |
2171 | radius = (1. - n(2)*n(2) - 4.*c*n(2)) / (4. * n(2) * n(2)); | |
2172 | ||
2173 | if (radius <= 0.E0) { | |
2174 | Error("RiemannFit", "Radius = %f less than zero!!!", radius); | |
2175 | return kFALSE; | |
2176 | } | |
2177 | radius = TMath::Sqrt(radius); | |
2178 | curv = 1.0 / radius; | |
2179 | ||
2180 | // evaluating signs for curvature and others | |
2181 | Double_t phi1 = 0.0, phi2, temp1, temp2, phi0, sumdphi = 0.0; | |
2182 | AliITSnode *p = fNode[0]; | |
2183 | phi1 = p->GetPhi(); | |
2184 | for (i = 1; i < dim; i++) { | |
2185 | p = (AliITSnode*)fNode[i]; | |
2186 | if (!p) break; | |
2187 | phi2 = p->GetPhi(); | |
2188 | temp1 = phi1; | |
2189 | temp2 = phi2; | |
2190 | if (temp1 > fgkPi && temp2 < fgkPi) | |
2191 | temp2 += fgkTwoPi; | |
2192 | else if (temp1 < fgkPi && temp2 > fgkPi) | |
2193 | temp1 += fgkTwoPi; | |
2194 | sumdphi += temp2 - temp1; | |
2195 | phi1 = phi2; | |
2196 | } | |
2197 | if (sumdphi < 0.E0) curv = -curv; | |
2198 | Double_t diff, angle = TMath::ATan2(yc, xc); | |
2199 | if (curv < 0.E0) | |
2200 | phi0 = angle + 0.5 * TMath::Pi(); | |
2201 | else | |
2202 | phi0 = angle - 0.5 * TMath::Pi(); | |
2203 | diff = angle - phi0; | |
2204 | ||
2205 | Double_t dt, temp = TMath::Sqrt(xc*xc + yc*yc) - radius; | |
2206 | if (curv >= 0.E0) | |
2207 | dt = temp; | |
2208 | else | |
2209 | dt = -temp; | |
2210 | //cout << "Dt = " << dt << endl; | |
2211 | ||
2212 | Double_t halfC = 0.5 * curv, test; | |
2213 | Double_t *s = new Double_t[dim], *zz = new Double_t[dim], *ws = new Double_t[dim]; | |
2214 | for (j = 0; j < 6; j++) { | |
2215 | p = fNode[j]; | |
2216 | if (!p) break; | |
2217 | //---- | |
2218 | s[j] = (p->GetR2sq() - dt * dt) / (1. + curv * dt); | |
2219 | if (s[j] < 0.) { | |
0db9364f | 2220 | if (TMath::Abs(s[j]) < 1.E-6) s[j] = 0.; |
cec46807 | 2221 | else { |
2222 | Error("RiemannFit", "Square root argument error: %17.15g < 0", s[j]); | |
2223 | return kFALSE; | |
2224 | } | |
2225 | } | |
2226 | s[j] = TMath::Sqrt(s[j]); | |
2227 | //cout << "Curv = " << halfC << " --- s[" << j << "] = " << s[j] << endl; | |
2228 | s[j] *= halfC; | |
2229 | test = TMath::Abs(s[j]); | |
2230 | if (test > 1.) { | |
2231 | if (test <= 1.1) | |
2232 | s[j] = ((s[j] > 0.) ? 0.99999999999 : -0.9999999999); | |
2233 | else { | |
2234 | Error("RiemannFit", "Value too large: %17.15g", s[j]); | |
2235 | return kFALSE; | |
2236 | } | |
2237 | } | |
2238 | //---- | |
2239 | zz[j] = p->Z(); | |
2240 | s[j] = TMath::ASin(s[j]) / halfC; | |
2241 | ws[j] = 1.0 / (p->ErrZ2()); | |
2242 | } | |
2243 | ||
2244 | // second tep final fit | |
2245 | Double_t s2Sum = 0.0, zSum = 0.0, szSum = 0.0, sSum = 0.0, sumw = 0.0; | |
2246 | for (i = 0; i < dim; i++) { | |
2247 | s2Sum += ws[i] * s[i] * s[i]; | |
2248 | zSum += ws[i] * zz[i]; | |
2249 | sSum += ws[i] * s[i]; | |
2250 | szSum += ws[i] * s[i] * zz[i]; | |
2251 | sumw += ws[i]; | |
2252 | } | |
2253 | s2Sum /= sumw; | |
2254 | zSum /= sumw; | |
2255 | sSum /= sumw; | |
2256 | szSum /= sumw; | |
2257 | temp = s2Sum - sSum*sSum; | |
2258 | ||
2259 | Double_t dz, tanL; | |
2260 | dz = (s2Sum*zSum - sSum*szSum) / temp; | |
2261 | tanL = (szSum - sSum*zSum) / temp; | |
2262 | ||
2263 | fXCenter = xc; | |
2264 | fYCenter = yc; | |
2265 | fRadius = radius; | |
2266 | fCurv = curv; | |
2267 | fPhi = phi0; | |
2268 | fDTrans = dt; | |
2269 | fDLong = dz; | |
2270 | fTanLambda = tanL; | |
2271 | ||
2272 | delete [] s; | |
2273 | delete [] zz; | |
2274 | delete [] ws; | |
2275 | ||
2276 | return kTRUE; | |
2277 | } |