-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathTMVA_MLP_Forward1stLoop.h
591 lines (538 loc) · 25.5 KB
/
TMVA_MLP_Forward1stLoop.h
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
// Class: ReadMLP_Forward1stLoop
// Automatically generated by MethodBase::MakeClass
//
/* configuration options =====================================================
#GEN -*-*-*-*-*-*-*-*-*-*-*- general info -*-*-*-*-*-*-*-*-*-*-*-
Method : MLP::MLP_Forward1stLoop
TMVA Release : 4.2.1 [262657]
ROOT Release : 6.06/00 [394752]
Creator : tnikodem
Date : Thu May 12 23:31:40 2016
Host : Linux lcgapp-slc6-physical1.cern.ch 2.6.32-573.8.1.el6.x86_64 #1 SMP Wed Nov 11 15:27:45 CET 2015 x86_64 x86_64 x86_64 GNU/Linux
Dir : /afs/cern.ch/work/t/tnikodem/work/FwdFitParams2/NeedforSpeed_Forward/Brunel_v49r1/runBrunel/TMVA
Training events: 1331762
Analysis type : [Classification]
#OPT -*-*-*-*-*-*-*-*-*-*-*-*- options -*-*-*-*-*-*-*-*-*-*-*-*-
# Set by User:
NCycles: "1500" [Number of training cycles]
HiddenLayers: "N+3,N+2,N" [Specification of hidden layer architecture]
NeuronType: "ReLU" [Neuron activation function type]
EstimatorType: "CE" [MSE (Mean Square Estimator) for Gaussian Likelihood or CE(Cross-Entropy) for Bernoulli Likelihood]
V: "False" [Verbose output (short form of "VerbosityLevel" below - overrides the latter one)]
VarTransform: "N" [List of variable transformations performed before training, e.g., "D_Background,P_Signal,G,N_AllClasses" for: "Decorrelation, PCA-transformation, Gaussianisation, Normalisation, each for the given class of events ('AllClasses' denotes all events of all classes, if no class indication is given, 'All' is assumed)"]
H: "True" [Print method-specific help message]
LearningRate: "1.000000e-02" [ANN learning rate parameter]
DecayRate: "5.000000e-03" [Decay rate for learning parameter]
TestRate: "5" [Test for overtraining performed at each #th epochs]
UseRegulator: "False" [Use regulator to avoid over-training]
# Default:
RandomSeed: "1" [Random seed for initial synapse weights (0 means unique seed for each run; default value '1')]
NeuronInputType: "sum" [Neuron input function type]
VerbosityLevel: "Default" [Verbosity level]
CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)]
IgnoreNegWeightsInTraining: "False" [Events with negative weights are ignored in the training (but are included for testing and performance evaluation)]
TrainingMethod: "BP" [Train with Back-Propagation (BP), BFGS Algorithm (BFGS), or Genetic Algorithm (GA - slower and worse)]
EpochMonitoring: "False" [Provide epoch-wise monitoring plots according to TestRate (caution: causes big ROOT output file!)]
Sampling: "1.000000e+00" [Only 'Sampling' (randomly selected) events are trained each epoch]
SamplingEpoch: "1.000000e+00" [Sampling is used for the first 'SamplingEpoch' epochs, afterwards, all events are taken for training]
SamplingImportance: "1.000000e+00" [ The sampling weights of events in epochs which successful (worse estimator than before) are multiplied with SamplingImportance, else they are divided.]
SamplingTraining: "True" [The training sample is sampled]
SamplingTesting: "False" [The testing sample is sampled]
ResetStep: "50" [How often BFGS should reset history]
Tau: "3.000000e+00" [LineSearch "size step"]
BPMode: "sequential" [Back-propagation learning mode: sequential or batch]
BatchSize: "-1" [Batch size: number of events/batch, only set if in Batch Mode, -1 for BatchSize=number_of_events]
ConvergenceImprove: "1.000000e-30" [Minimum improvement which counts as improvement (<0 means automatic convergence check is turned off)]
ConvergenceTests: "-1" [Number of steps (without improvement) required for convergence (<0 means automatic convergence check is turned off)]
UpdateLimit: "10000" [Maximum times of regulator update]
CalculateErrors: "False" [Calculates inverse Hessian matrix at the end of the training to be able to calculate the uncertainties of an MVA value]
WeightRange: "1.000000e+00" [Take the events for the estimator calculations from small deviations from the desired value to large deviations only over the weight range]
##
#VAR -*-*-*-*-*-*-*-*-*-*-*-* variables *-*-*-*-*-*-*-*-*-*-*-*-
NVar 7
nPlanes nPlanes nPlanes nPlanes 'I' [10,12]
dSlope p p p 'F' [-0.000876020232681,0.000809539866168]
dp dp dp dp 'F' [-0.000959908822551,0.000927863351535]
slope2 slope2 slope2 slope2 'F' [6.83568359818e-05,0.171713337302]
dby dby dby dby 'F' [-0.502270042896,0.598416030407]
dbx dbx dbx dbx 'F' [-0.0183157920837,0.0351094007492]
day day day day 'F' [-178.507324219,188.444091797]
NSpec 0
============================================================================ */
#include <vector>
#include <cmath>
#include <string>
#include <iostream>
#ifndef IClassifierReader__def
#define IClassifierReader__def
class IClassifierReader {
public:
// constructor
IClassifierReader() {}
virtual ~IClassifierReader() {}
// return classifier response
virtual float GetMvaValue( std::vector<float>& inputValues ) const = 0;
};
#endif
class ReadMLP_Forward1stLoop : public IClassifierReader {
public:
// constructor
ReadMLP_Forward1stLoop( const std::vector<std::string>& theInputVars )
: IClassifierReader(),
fClassName( "ReadMLP_Forward1stLoop" ),
fNvars( 7 )
{
// the training input variables
const char* inputVars[] = { "nPlanes", "dSlope", "dp", "slope2", "dby", "dbx", "day" };
// sanity checks
if (theInputVars.size() <= 0) {
std::cout << "Problem in class \"" << fClassName << "\": empty input vector" << std::endl;
return;
}
if (theInputVars.size() != fNvars) {
std::cout << "Problem in class \"" << fClassName << "\": mismatch in number of input values: "
<< theInputVars.size() << " != " << fNvars << std::endl;
return;
}
// validate input variables
for (size_t ivar = 0; ivar < theInputVars.size(); ivar++) {
if (theInputVars[ivar] != inputVars[ivar]) {
std::cout << "Problem in class \"" << fClassName << "\": mismatch in input variable names" << std::endl
<< " for variable [" << ivar << "]: " << theInputVars[ivar].c_str() << " != " << inputVars[ivar] << std::endl;
return;
}
}
// initialize constants
Initialize();
// initialize transformation
InitTransform_1();
}
// destructor
virtual ~ReadMLP_Forward1stLoop() {
}
// the classifier response
// "inputValues" is a vector of input values in the same order as the
// variables given to the constructor
// WARNING: inputVariables will be modified
inline float GetMvaValue( std::vector<float>& iV ) const override
{
//Normalize input
Transform_1( iV);
return GetMvaValue__( iV );
}
private:
// input variable transformation
float fMin_1[3][7];
float fMax_1[3][7];
inline void InitTransform_1()
{
// Normalization transformation, initialisation
fMin_1[0][0] = 10;
fMax_1[0][0] = 12;
fMin_1[1][0] = 10;
fMax_1[1][0] = 12;
fMin_1[2][0] = 10;
fMax_1[2][0] = 12;
fMin_1[0][1] = -0.000693323207088;
fMax_1[0][1] = 0.000661431695335;
fMin_1[1][1] = -0.000876020232681;
fMax_1[1][1] = 0.000809539866168;
fMin_1[2][1] = -0.000876020232681;
fMax_1[2][1] = 0.000809539866168;
fMin_1[0][2] = -0.000726951810066;
fMax_1[0][2] = 0.000800277863164;
fMin_1[1][2] = -0.000959908822551;
fMax_1[1][2] = 0.000927863351535;
fMin_1[2][2] = -0.000959908822551;
fMax_1[2][2] = 0.000927863351535;
fMin_1[0][3] = 8.73411772773e-05;
fMax_1[0][3] = 0.161471515894;
fMin_1[1][3] = 6.83568359818e-05;
fMax_1[1][3] = 0.171713337302;
fMin_1[2][3] = 6.83568359818e-05;
fMax_1[2][3] = 0.171713337302;
fMin_1[0][4] = -0.41890129447;
fMax_1[0][4] = 0.25404638052;
fMin_1[1][4] = -0.502270042896;
fMax_1[1][4] = 0.598416030407;
fMin_1[2][4] = -0.502270042896;
fMax_1[2][4] = 0.598416030407;
fMin_1[0][5] = -0.0150038003922;
fMax_1[0][5] = 0.0172826647758;
fMin_1[1][5] = -0.0183157920837;
fMax_1[1][5] = 0.0351094007492;
fMin_1[2][5] = -0.0183157920837;
fMax_1[2][5] = 0.0351094007492;
fMin_1[0][6] = -139.267822266;
fMax_1[0][6] = 150.771850586;
fMin_1[1][6] = -178.507324219;
fMax_1[1][6] = 188.444091797;
fMin_1[2][6] = -178.507324219;
fMax_1[2][6] = 188.444091797;
}
inline void Transform_1( std::vector<float>& iv) const
{
const int cls = 2; //what are the other???
const int nVar = 7;
for (int ivar=0;ivar<nVar;ivar++) {
const float offset = fMin_1[cls][ivar];
const float scale = 1.0/(fMax_1[cls][ivar]-fMin_1[cls][ivar]); //TODO speed this up. but then not easy to update :(
iv[ivar] = (iv[ivar]-offset)*scale * 2. - 1.;
}
}
// common member variables
const char* fClassName;
const size_t fNvars;
// initialize internal variables
inline void Initialize()
{
// build network structure
fLayers = 5;
fLayerSize[0] = 8;
fLayerSize[1] = 11;
fLayerSize[2] = 10;
fLayerSize[3] = 8;
fLayerSize[4] = 1;
// weight matrix from layer 0 to 1
fWeightMatrix0to1[0][0] = 0.156511290357907;
fWeightMatrix0to1[1][0] = 5.859677434496;
fWeightMatrix0to1[2][0] = 0.0403216550547087;
fWeightMatrix0to1[3][0] = 0.118942942089809;
fWeightMatrix0to1[4][0] = -0.0524130927108231;
fWeightMatrix0to1[5][0] = -0.0615780376417546;
fWeightMatrix0to1[6][0] = 0.100115445436162;
fWeightMatrix0to1[7][0] = 0.0612125106565374;
fWeightMatrix0to1[8][0] = 0.0351187781245128;
fWeightMatrix0to1[9][0] = -7.80295272542051;
fWeightMatrix0to1[0][1] = -15.9012468978926;
fWeightMatrix0to1[1][1] = 0.103375966424631;
fWeightMatrix0to1[2][1] = 2.84181299733342;
fWeightMatrix0to1[3][1] = 14.0872677784526;
fWeightMatrix0to1[4][1] = 0.00656033420416158;
fWeightMatrix0to1[5][1] = -2.05049322142208;
fWeightMatrix0to1[6][1] = -0.476796624482398;
fWeightMatrix0to1[7][1] = 5.34178801874048;
fWeightMatrix0to1[8][1] = -2.03842114206132;
fWeightMatrix0to1[9][1] = 1.61796425984333;
fWeightMatrix0to1[0][2] = -0.626716696225122;
fWeightMatrix0to1[1][2] = 0.293994980752188;
fWeightMatrix0to1[2][2] = -5.8378296035011;
fWeightMatrix0to1[3][2] = 1.88698468349361;
fWeightMatrix0to1[4][2] = 2.59126699636858;
fWeightMatrix0to1[5][2] = 2.76426352177771;
fWeightMatrix0to1[6][2] = 35.7389195659674;
fWeightMatrix0to1[7][2] = 15.3804699396489;
fWeightMatrix0to1[8][2] = -34.4263535379079;
fWeightMatrix0to1[9][2] = 0.217589935429331;
fWeightMatrix0to1[0][3] = 5.04491816907169;
fWeightMatrix0to1[1][3] = 4.99503349862526;
fWeightMatrix0to1[2][3] = -9.72920646551164;
fWeightMatrix0to1[3][3] = 7.30206859198975;
fWeightMatrix0to1[4][3] = -3.63707417363857;
fWeightMatrix0to1[5][3] = -5.86739614942788;
fWeightMatrix0to1[6][3] = -4.13376983329174;
fWeightMatrix0to1[7][3] = -8.62681866836684;
fWeightMatrix0to1[8][3] = -2.35484344660064;
fWeightMatrix0to1[9][3] = 15.6094338591807;
fWeightMatrix0to1[0][4] = -4.80684731442771;
fWeightMatrix0to1[1][4] = 1.42397639088434;
fWeightMatrix0to1[2][4] = 18.6262099897472;
fWeightMatrix0to1[3][4] = -1.30290028722016;
fWeightMatrix0to1[4][4] = -49.5975433863944;
fWeightMatrix0to1[5][4] = -24.9254684302377;
fWeightMatrix0to1[6][4] = 30.3398826436804;
fWeightMatrix0to1[7][4] = -1.40266667371555;
fWeightMatrix0to1[8][4] = 33.3763741231589;
fWeightMatrix0to1[9][4] = 2.83709336083157;
fWeightMatrix0to1[0][5] = 2.16633228579812;
fWeightMatrix0to1[1][5] = 0.149316763261451;
fWeightMatrix0to1[2][5] = -28.0510120484557;
fWeightMatrix0to1[3][5] = 1.48947030290536;
fWeightMatrix0to1[4][5] = 3.31909225272089;
fWeightMatrix0to1[5][5] = 1.50257870289561;
fWeightMatrix0to1[6][5] = -6.42093408725713;
fWeightMatrix0to1[7][5] = -4.01613499600723;
fWeightMatrix0to1[8][5] = 22.5853569887894;
fWeightMatrix0to1[9][5] = -0.988703229599317;
fWeightMatrix0to1[0][6] = -0.515475069093284;
fWeightMatrix0to1[1][6] = 0.107873491456676;
fWeightMatrix0to1[2][6] = 3.7896539966571;
fWeightMatrix0to1[3][6] = -1.92206830502861;
fWeightMatrix0to1[4][6] = 18.4492434520554;
fWeightMatrix0to1[5][6] = -31.3178255977452;
fWeightMatrix0to1[6][6] = 4.37461684624034;
fWeightMatrix0to1[7][6] = 3.42005751481484;
fWeightMatrix0to1[8][6] = -7.52309200432279;
fWeightMatrix0to1[9][6] = 1.16121311781456;
fWeightMatrix0to1[0][7] = 5.69099525042923;
fWeightMatrix0to1[1][7] = -0.709077849551806;
fWeightMatrix0to1[2][7] = -16.3792775664682;
fWeightMatrix0to1[3][7] = 6.82749219677059;
fWeightMatrix0to1[4][7] = -5.68770379136056;
fWeightMatrix0to1[5][7] = -8.26036569859038;
fWeightMatrix0to1[6][7] = -3.79213899618358;
fWeightMatrix0to1[7][7] = -10.6438062975721;
fWeightMatrix0to1[8][7] = 8.20014457995756;
fWeightMatrix0to1[9][7] = 8.20255329829369;
// weight matrix from layer 1 to 2
fWeightMatrix1to2[0][0] = -1.12084468217478;
fWeightMatrix1to2[1][0] = -1.42097522295155;
fWeightMatrix1to2[2][0] = 0.0782684537101741;
fWeightMatrix1to2[3][0] = -0.975849184350193;
fWeightMatrix1to2[4][0] = -1.36256041673412;
fWeightMatrix1to2[5][0] = -0.263140244746088;
fWeightMatrix1to2[6][0] = -1.21197406229234;
fWeightMatrix1to2[7][0] = 0.693466224402045;
fWeightMatrix1to2[8][0] = 1.08487246925391;
fWeightMatrix1to2[0][1] = -0.0446782697496722;
fWeightMatrix1to2[1][1] = -1.67332132886065;
fWeightMatrix1to2[2][1] = -0.935966205565634;
fWeightMatrix1to2[3][1] = -0.711903382557495;
fWeightMatrix1to2[4][1] = -2.52251525746623;
fWeightMatrix1to2[5][1] = -2.53192420411418;
fWeightMatrix1to2[6][1] = 0.0831796043408627;
fWeightMatrix1to2[7][1] = -2.78890019799941;
fWeightMatrix1to2[8][1] = 1.03799008336373;
fWeightMatrix1to2[0][2] = -0.224480195037057;
fWeightMatrix1to2[1][2] = 0.610785468341552;
fWeightMatrix1to2[2][2] = -2.92188801884268;
fWeightMatrix1to2[3][2] = -0.774161656278927;
fWeightMatrix1to2[4][2] = 0.29591419521103;
fWeightMatrix1to2[5][2] = 0.160621529223216;
fWeightMatrix1to2[6][2] = -7.28313581268633;
fWeightMatrix1to2[7][2] = 0.896688799158805;
fWeightMatrix1to2[8][2] = -0.255417479930132;
fWeightMatrix1to2[0][3] = -1.17399857513146;
fWeightMatrix1to2[1][3] = -1.49788780317135;
fWeightMatrix1to2[2][3] = -2.85320287621488;
fWeightMatrix1to2[3][3] = -0.654462509613729;
fWeightMatrix1to2[4][3] = -0.460201028616304;
fWeightMatrix1to2[5][3] = -0.91892110249576;
fWeightMatrix1to2[6][3] = 0.408582299903928;
fWeightMatrix1to2[7][3] = -2.13617351122212;
fWeightMatrix1to2[8][3] = 0.846400463160404;
fWeightMatrix1to2[0][4] = 1.18274157694714;
fWeightMatrix1to2[1][4] = 0.363577818975712;
fWeightMatrix1to2[2][4] = -2.78648793907969;
fWeightMatrix1to2[3][4] = -0.605550298034862;
fWeightMatrix1to2[4][4] = -3.56776391965204;
fWeightMatrix1to2[5][4] = -0.824338706374845;
fWeightMatrix1to2[6][4] = -0.0739282265369494;
fWeightMatrix1to2[7][4] = -0.617919855531056;
fWeightMatrix1to2[8][4] = -1.26441402791213;
fWeightMatrix1to2[0][5] = -7.85422121839153;
fWeightMatrix1to2[1][5] = 0.649212928279747;
fWeightMatrix1to2[2][5] = -1.49382735672959;
fWeightMatrix1to2[3][5] = -1.163806181279;
fWeightMatrix1to2[4][5] = -0.706737257250919;
fWeightMatrix1to2[5][5] = -0.320001669951044;
fWeightMatrix1to2[6][5] = -0.915244188961159;
fWeightMatrix1to2[7][5] = -0.404929788423475;
fWeightMatrix1to2[8][5] = -0.869837982881825;
fWeightMatrix1to2[0][6] = 0.388461702018911;
fWeightMatrix1to2[1][6] = 0.82417411752276;
fWeightMatrix1to2[2][6] = -1.16986480480305;
fWeightMatrix1to2[3][6] = -0.594234505980216;
fWeightMatrix1to2[4][6] = 1.01190529478994;
fWeightMatrix1to2[5][6] = 0.488157766444359;
fWeightMatrix1to2[6][6] = 1.21927172754306;
fWeightMatrix1to2[7][6] = 1.03555206522933;
fWeightMatrix1to2[8][6] = -0.432120285066959;
fWeightMatrix1to2[0][7] = -0.720718150886782;
fWeightMatrix1to2[1][7] = -0.772587109315925;
fWeightMatrix1to2[2][7] = 4.961492917941;
fWeightMatrix1to2[3][7] = -2.07551955936462;
fWeightMatrix1to2[4][7] = -1.49050888933505;
fWeightMatrix1to2[5][7] = -5.18104528023642;
fWeightMatrix1to2[6][7] = -3.41535949506006;
fWeightMatrix1to2[7][7] = -3.63937612748511;
fWeightMatrix1to2[8][7] = -0.0969639864582296;
fWeightMatrix1to2[0][8] = 0.118654098736673;
fWeightMatrix1to2[1][8] = 0.643045299775251;
fWeightMatrix1to2[2][8] = -6.68887454121865;
fWeightMatrix1to2[3][8] = -0.844473100072775;
fWeightMatrix1to2[4][8] = 1.65222133238673;
fWeightMatrix1to2[5][8] = 0.770190997903331;
fWeightMatrix1to2[6][8] = 1.28375929668021;
fWeightMatrix1to2[7][8] = -0.00194208275099416;
fWeightMatrix1to2[8][8] = -0.531798783461306;
fWeightMatrix1to2[0][9] = 0.293772101944011;
fWeightMatrix1to2[1][9] = 0.476087002004467;
fWeightMatrix1to2[2][9] = -2.86339487514251;
fWeightMatrix1to2[3][9] = -0.106973459039663;
fWeightMatrix1to2[4][9] = 0.237379536635383;
fWeightMatrix1to2[5][9] = -6.55306519686434;
fWeightMatrix1to2[6][9] = 0.205897405036469;
fWeightMatrix1to2[7][9] = 0.795064632110721;
fWeightMatrix1to2[8][9] = -0.962009195010732;
fWeightMatrix1to2[0][10] = 2.4000326297059;
fWeightMatrix1to2[1][10] = 3.79579237455207;
fWeightMatrix1to2[2][10] = 7.69085292477427;
fWeightMatrix1to2[3][10] = 7.36111070439386;
fWeightMatrix1to2[4][10] = 1.85118194058879;
fWeightMatrix1to2[5][10] = -0.0521258745313225;
fWeightMatrix1to2[6][10] = 1.83688685910519;
fWeightMatrix1to2[7][10] = -2.36525257885871;
fWeightMatrix1to2[8][10] = 3.76587549476592;
// weight matrix from layer 2 to 3
fWeightMatrix2to3[0][0] = -1.16975284882838;
fWeightMatrix2to3[1][0] = 0.182436124863262;
fWeightMatrix2to3[2][0] = 1.32004139776896;
fWeightMatrix2to3[3][0] = -0.813332692867775;
fWeightMatrix2to3[4][0] = -0.407974392591066;
fWeightMatrix2to3[5][0] = -0.298095908286076;
fWeightMatrix2to3[6][0] = -0.209551158952576;
fWeightMatrix2to3[0][1] = 0.0591791985821578;
fWeightMatrix2to3[1][1] = 0.496981191089355;
fWeightMatrix2to3[2][1] = 0.125152050979222;
fWeightMatrix2to3[3][1] = -1.06495030957181;
fWeightMatrix2to3[4][1] = -0.548374906025986;
fWeightMatrix2to3[5][1] = -0.124331842255717;
fWeightMatrix2to3[6][1] = 1.50571085321915;
fWeightMatrix2to3[0][2] = -0.50305998401584;
fWeightMatrix2to3[1][2] = 1.11074917712974;
fWeightMatrix2to3[2][2] = -0.187830512893655;
fWeightMatrix2to3[3][2] = -0.796139378341234;
fWeightMatrix2to3[4][2] = 0.156072974019272;
fWeightMatrix2to3[5][2] = -0.0467995879623844;
fWeightMatrix2to3[6][2] = 0.0265475889574797;
fWeightMatrix2to3[0][3] = 0.180573513036599;
fWeightMatrix2to3[1][3] = -0.392574059691424;
fWeightMatrix2to3[2][3] = -0.56129662210301;
fWeightMatrix2to3[3][3] = 1.39644020937997;
fWeightMatrix2to3[4][3] = 3.17536266287918;
fWeightMatrix2to3[5][3] = 0.462627410049255;
fWeightMatrix2to3[6][3] = -0.271723558819108;
fWeightMatrix2to3[0][4] = 0.281470936587779;
fWeightMatrix2to3[1][4] = -3.07088331788158;
fWeightMatrix2to3[2][4] = -0.0813497289253693;
fWeightMatrix2to3[3][4] = -0.347310938430066;
fWeightMatrix2to3[4][4] = -0.763993695187927;
fWeightMatrix2to3[5][4] = -1.94323588808135;
fWeightMatrix2to3[6][4] = -0.398015524174699;
fWeightMatrix2to3[0][5] = -0.91364566597847;
fWeightMatrix2to3[1][5] = -0.259606356782924;
fWeightMatrix2to3[2][5] = -1.25024846199983;
fWeightMatrix2to3[3][5] = 0.685295486731914;
fWeightMatrix2to3[4][5] = 2.66445157740408;
fWeightMatrix2to3[5][5] = -1.70863318714099;
fWeightMatrix2to3[6][5] = -0.56224569042501;
fWeightMatrix2to3[0][6] = -0.0781893585389807;
fWeightMatrix2to3[1][6] = -0.2157817435887;
fWeightMatrix2to3[2][6] = -0.166414327646281;
fWeightMatrix2to3[3][6] = -0.140533721077004;
fWeightMatrix2to3[4][6] = -0.401772332808343;
fWeightMatrix2to3[5][6] = -0.448814135432277;
fWeightMatrix2to3[6][6] = 0.585911073797476;
fWeightMatrix2to3[0][7] = -0.848544487609584;
fWeightMatrix2to3[1][7] = -0.673594428005596;
fWeightMatrix2to3[2][7] = -0.570735830368554;
fWeightMatrix2to3[3][7] = -0.337810117449569;
fWeightMatrix2to3[4][7] = -0.850964902420546;
fWeightMatrix2to3[5][7] = -1.16577487918861;
fWeightMatrix2to3[6][7] = 0.570034431102591;
fWeightMatrix2to3[0][8] = -0.701396529760941;
fWeightMatrix2to3[1][8] = 0.778845996680412;
fWeightMatrix2to3[2][8] = -0.313119262292289;
fWeightMatrix2to3[3][8] = 0.0538919388985925;
fWeightMatrix2to3[4][8] = 0.0837696630186691;
fWeightMatrix2to3[5][8] = -1.93455771407413;
fWeightMatrix2to3[6][8] = -0.0478752668570304;
fWeightMatrix2to3[0][9] = 3.78541028470798;
fWeightMatrix2to3[1][9] = -1.72163006143859;
fWeightMatrix2to3[2][9] = 0.603240734576226;
fWeightMatrix2to3[3][9] = 2.80666479501064;
fWeightMatrix2to3[4][9] = -11.6469071199572;
fWeightMatrix2to3[5][9] = -2.08478104540014;
fWeightMatrix2to3[6][9] = -3.80015882535704;
// weight matrix from layer 3 to 4
fWeightMatrix3to4[0][0] = -0.408291319489926;
fWeightMatrix3to4[0][1] = 0.433421843843804;
fWeightMatrix3to4[0][2] = -0.45970598445461;
fWeightMatrix3to4[0][3] = 0.531229831434917;
fWeightMatrix3to4[0][4] = 0.482734358752872;
fWeightMatrix3to4[0][5] = 0.975569090219759;
fWeightMatrix3to4[0][6] = -0.450251757645731;
fWeightMatrix3to4[0][7] = 0.4271524095352;
}
inline float GetMvaValue__( const std::vector<float>& inputValues ) const
{
//TODO check auto vectorization here: 'not vectorized: unsupported use in stmt' , 'Unsupported pattern'
//reply (Paul Seyfert): can be autovectorised with a few changes
// cf https://github.com/root-project/root/pull/121
// and https://github.com/pseyfert/tmva-mlp
float fWeights0[8];
float fWeights1[11];
float fWeights2[10];
float fWeights3[8];
float fWeights4[1];
for (int i=0; i<fLayerSize[0]; i++) fWeights0[i]=0.f;
for (int i=0; i<fLayerSize[1]; i++) fWeights1[i]=0.f;
for (int i=0; i<fLayerSize[2]; i++) fWeights2[i]=0.f;
for (int i=0; i<fLayerSize[3]; i++) fWeights3[i]=0.f;
for (int i=0; i<fLayerSize[4]; i++) fWeights4[i]=0.f;
fWeights0[fLayerSize[0]-1]=1.f;
fWeights1[fLayerSize[1]-1]=1.f;
fWeights2[fLayerSize[2]-1]=1.f;
fWeights3[fLayerSize[3]-1]=1.f;
for (int i=0; i<7; i++)
fWeights0[i]=inputValues[i];
// layer 0 to 1
float buffer[11];
for (int o=0; o<10; o++) {
for (int i=0; i<8; i++) {
buffer[i] = fWeightMatrix0to1[o][i] * fWeights0[i];
}
for (int i=0; i<8; i++) {
fWeights1[o] += buffer[i];
}
}
for (int o=0; o<10; o++) {
fWeights1[o] = ActivationFnc(fWeights1[o]);
}
// layer 1 to 2
for (int o=0; o<9; o++) {
for (int i=0; i<11; i++) {
buffer[i] = fWeightMatrix1to2[o][i] * fWeights1[i];
}
for (int i=0; i<11; i++) {
fWeights2[o] += buffer[i];
}
fWeights2[o] = ActivationFnc(fWeights2[o]);
}
// layer 2 to 3
for (int o=0; o<7; o++) {
for (int i=0; i<10; i++) {
buffer[i] = fWeightMatrix2to3[o][i] * fWeights2[i];
}
for (int i=0; i<10; i++) {
fWeights3[o] += buffer[i];
}
fWeights3[o] = ActivationFnc(fWeights3[o]);
}
// layer 3 to 4
for (int i=0; i<8; i++) {
buffer[i] = fWeightMatrix3to4[0][i] * fWeights3[i];
}
for (int i=0; i<8; i++) {
fWeights4[0] += buffer[i];
}
return OutputActivationFnc(fWeights4[0]);
}
inline float ActivationFnc(float x) const {
// rectified linear unit
return x*(x>0);
}
//TODO do we need exp here? can we live with pure x output?!
//reply (Paul Seyfert): this transformation is indeed not needed (provided the cut value gets adapted)
inline float OutputActivationFnc(float x) const {
// sigmoid
return 1.0/(1.0+exp(-x));
}
int fLayers;
int fLayerSize[5];
float fWeightMatrix0to1[11][8]; // weight matrix from layer 0 to 1
float fWeightMatrix1to2[10][11]; // weight matrix from layer 1 to 2
float fWeightMatrix2to3[8][10]; // weight matrix from layer 2 to 3
float fWeightMatrix3to4[1][8]; // weight matrix from layer 3 to 4
};