Execution Time0.49s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: PR-4624-x86_64-mac1014-clang100-opt (macitois21.cern.ch) on 2019-11-14 18:12:19

Test Timing: Passed
Processors1

Show Command Line
Display graphs:

Test output
Testing Backpropagation:
DEEP NEURAL NETWORK:   Depth = 3  Input = ( 1, 10, 4 )  Batch size = 10  Loss function = R
	Layer 0	 DENSE Layer: 	 ( Input =     4 , Width =     2 ) 	Output = (  1 ,    10 ,     2 ) 	 Activation Function = Identity
	Layer 1	 BATCH NORM Layer: 	 ( Input =     2 ) 
	Layer 2	 DENSE Layer: 	 ( Input =     2 , Width =     1 ) 	Output = (  1 ,    10 ,     1 ) 	 Activation Function = Identity
input 

10x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |     0.9989     -0.4348      0.7818    -0.03005 
   1 |     0.8243    -0.05672     -0.9009     -0.0747 
   2 |   0.007912     -0.4108       1.391     -0.9851 
   3 |   -0.04894      -1.443      -1.061      -1.388 
   4 |     0.7674      -0.736      0.5797     -0.3821 
   5 |      2.061      -1.235       1.165     -0.4542 
   6 |    -0.1348     -0.4996     -0.1824       1.844 
   7 |    -0.2428       1.997    0.004806     -0.4222 
   8 |      1.541     0.09474       1.525       1.217 
   9 |    -0.1363     -0.1992     -0.2938     -0.1184 

 training batch 1 mu var0-0.237575
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.7362     -0.8164 
   1 |     0.7832       1.481 
   2 |    -0.4308      -1.326 
   3 |      2.162         1.9 
   4 |     -0.231     -0.4749 
   5 |    -0.7836     -0.9299 
   6 |     -1.232       -1.08 
   7 |     0.1847      0.6814 
   8 |     -2.477      -2.282 
   9 |      0.385      0.3715 

output BN 
output DL feature 0 mean -0.237575	output DL std 1.24516
output DL feature 1 mean -0.247568	output DL std 1.32001
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.4221     -0.4542 
   1 |     0.8641        1.38 
   2 |    -0.1636     -0.8612 
   3 |      2.031       1.715 
   4 |   0.005564     -0.1815 
   5 |    -0.4622     -0.5448 
   6 |    -0.8421     -0.6649 
   7 |     0.3575      0.7418 
   8 |     -1.896      -1.625 
   9 |      0.527      0.4943 

output BN feature 0 mean 3.33067e-17	output BN std 1.05405
output BN feature 1 mean 2.22045e-17	output BN std 1.05406
Testing weight gradients   for    layer 0
weight gradient for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |  -0.009419    -0.08376     0.06973    -0.06887 
   1 |    0.05391     0.07895     0.03514      0.1341 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |    -0.1001    -0.08378     -0.8906     -0.7863 
   1 |     0.3094      0.2496      -1.325     -0.6265 

 training batch 2 mu var0-0.237572
compute loss for weight  -0.100132  -0.100142 result 0.0385235
 training batch 3 mu var0-0.237575
compute loss for weight  -0.100152  -0.100142 result 0.0385237
 training batch 4 mu var0-0.237575
compute loss for weight  -0.100137  -0.100142 result 0.0385235
 training batch 5 mu var0-0.237575
compute loss for weight  -0.100147  -0.100142 result 0.0385236
   --dy = -0.00941944 dy_ref = -0.00941944
 training batch 6 mu var0-0.237576
compute loss for weight  -0.0837655  -0.0837755 result 0.0385227
 training batch 7 mu var0-0.237575
compute loss for weight  -0.0837855  -0.0837755 result 0.0385244
 training batch 8 mu var0-0.237575
compute loss for weight  -0.0837705  -0.0837755 result 0.0385232
 training batch 9 mu var0-0.237575
compute loss for weight  -0.0837805  -0.0837755 result 0.038524
   --dy = -0.0837587 dy_ref = -0.0837587
 training batch 10 mu var0-0.237575
compute loss for weight  -0.890559  -0.890569 result 0.0385243
 training batch 11 mu var0-0.237575
compute loss for weight  -0.890579  -0.890569 result 0.0385229
 training batch 12 mu var0-0.237575
compute loss for weight  -0.890564  -0.890569 result 0.0385239
 training batch 13 mu var0-0.237575
compute loss for weight  -0.890574  -0.890569 result 0.0385232
   --dy = 0.0697326 dy_ref = 0.0697326
 training batch 14 mu var0-0.237575
compute loss for weight  -0.786277  -0.786287 result 0.0385229
 training batch 15 mu var0-0.237575
compute loss for weight  -0.786297  -0.786287 result 0.0385243
 training batch 16 mu var0-0.237575
compute loss for weight  -0.786282  -0.786287 result 0.0385232
 training batch 17 mu var0-0.237575
compute loss for weight  -0.786292  -0.786287 result 0.0385239
   --dy = -0.0688663 dy_ref = -0.0688663
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    0.09986    -0.02281 

weights for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |          1           1 

 training batch 18 mu var0-0.237575
compute loss for weight  1.00001  1 result 0.0385246
 training batch 19 mu var0-0.237575
compute loss for weight  0.99999  1 result 0.0385226
 training batch 20 mu var0-0.237575
compute loss for weight  1.00001  1 result 0.0385241
 training batch 21 mu var0-0.237575
compute loss for weight  0.999995  1 result 0.0385231
   --dy = 0.0998619 dy_ref = 0.0998619
 training batch 22 mu var0-0.237575
compute loss for weight  1.00001  1 result 0.0385233
 training batch 23 mu var0-0.237575
compute loss for weight  0.99999  1 result 0.0385238
 training batch 24 mu var0-0.237575
compute loss for weight  1.00001  1 result 0.0385235
 training batch 25 mu var0-0.237575
compute loss for weight  0.999995  1 result 0.0385237
   --dy = -0.0228147 dy_ref = -0.0228147
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 | -3.469e-18  -1.735e-18 

weights for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |          0           0 

 training batch 26 mu var0-0.237575
compute loss for weight  1e-05  0 result 0.0385236
 training batch 27 mu var0-0.237575
compute loss for weight  -1e-05  0 result 0.0385236
 training batch 28 mu var0-0.237575
compute loss for weight  5e-06  0 result 0.0385236
 training batch 29 mu var0-0.237575
compute loss for weight  -5e-06  0 result 0.0385236
   --dy = -1.96602e-12 dy_ref = -3.46945e-18
 training batch 30 mu var0-0.237575
compute loss for weight  1e-05  0 result 0.0385236
 training batch 31 mu var0-0.237575
compute loss for weight  -1e-05  0 result 0.0385236
 training batch 32 mu var0-0.237575
compute loss for weight  5e-06  0 result 0.0385236
 training batch 33 mu var0-0.237575
compute loss for weight  -5e-06  0 result 0.0385236
   --dy = -1.04083e-12 dy_ref = -1.73472e-18
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.1711     0.04316 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.5837     -0.5286 

 training batch 34 mu var0-0.237575
compute loss for weight  0.583721  0.583711 result 0.0385253
 training batch 35 mu var0-0.237575
compute loss for weight  0.583701  0.583711 result 0.0385219
 training batch 36 mu var0-0.237575
compute loss for weight  0.583716  0.583711 result 0.0385244
 training batch 37 mu var0-0.237575
compute loss for weight  0.583706  0.583711 result 0.0385227
   --dy = 0.171081 dy_ref = 0.171081
 training batch 38 mu var0-0.237575
compute loss for weight  -0.528549  -0.528559 result 0.038524
 training batch 39 mu var0-0.237575
compute loss for weight  -0.528569  -0.528559 result 0.0385231
 training batch 40 mu var0-0.237575
compute loss for weight  -0.528554  -0.528559 result 0.0385238
 training batch 41 mu var0-0.237575
compute loss for weight  -0.528564  -0.528559 result 0.0385234
   --dy = 0.0431641 dy_ref = 0.0431641
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m6.78355e-10[NON-XML-CHAR-0x1B][39m