Execution Time0.56s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: PR-4623-x86_64-fedora30-gcc9-opt (root-fedora30-1.cern.ch) on 2019-11-14 10:10:32

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 var0-0.534827
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     -1.055     -0.4848 
   1 |   -0.06133      0.2991 
   2 |    -0.3885      -1.204 
   3 |     0.3244     -0.9429 
   4 |    -0.8527     -0.7253 
   5 |     -1.997      -1.178 
   6 |    -0.8299      0.7558 
   7 |       1.26      0.6882 
   8 |     -1.894      0.1059 
   9 |     0.1458    -0.04092 

output BN 
output DL feature 0 mean -0.534827	output DL std 1.00638
output DL feature 1 mean -0.272709	output DL std 0.737221
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.5449     -0.3032 
   1 |     0.4959      0.8175 
   2 |     0.1533      -1.331 
   3 |     0.8999     -0.9582 
   4 |     -0.333      -0.647 
   5 |     -1.531      -1.295 
   6 |     -0.309        1.47 
   7 |       1.88       1.374 
   8 |     -1.424      0.5413 
   9 |     0.7129      0.3314 

output BN feature 0 mean -1.11022e-17	output BN std 1.05403
output BN feature 1 mean -6.66134e-17	output BN std 1.05398
Testing weight gradients   for    layer 0
weight gradient for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |   0.007502    -0.08054     0.02258      -0.131 
   1 |      0.133    -0.06559      0.1181      0.1374 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |    -0.5322      0.4831     -0.4162     -0.3991 
   1 |    0.02416      0.4535     -0.3797      0.4969 

 training batch 2 mu var0-0.534825
compute loss for weight  -0.532209  -0.532219 result 0.218525
 training batch 3 mu var0-0.534827
compute loss for weight  -0.532229  -0.532219 result 0.218524
 training batch 4 mu var0-0.534827
compute loss for weight  -0.532214  -0.532219 result 0.218525
 training batch 5 mu var0-0.534827
compute loss for weight  -0.532224  -0.532219 result 0.218524
   --dy = 0.00750179 dy_ref = 0.00750179
 training batch 6 mu var0-0.534828
compute loss for weight  0.483079  0.483069 result 0.218524
 training batch 7 mu var0-0.534827
compute loss for weight  0.483059  0.483069 result 0.218525
 training batch 8 mu var0-0.534828
compute loss for weight  0.483074  0.483069 result 0.218524
 training batch 9 mu var0-0.534827
compute loss for weight  0.483064  0.483069 result 0.218525
   --dy = -0.0805399 dy_ref = -0.0805399
 training batch 10 mu var0-0.534827
compute loss for weight  -0.416188  -0.416198 result 0.218525
 training batch 11 mu var0-0.534827
compute loss for weight  -0.416208  -0.416198 result 0.218524
 training batch 12 mu var0-0.534827
compute loss for weight  -0.416193  -0.416198 result 0.218525
 training batch 13 mu var0-0.534827
compute loss for weight  -0.416203  -0.416198 result 0.218524
   --dy = 0.0225831 dy_ref = 0.0225831
 training batch 14 mu var0-0.534827
compute loss for weight  -0.399135  -0.399145 result 0.218523
 training batch 15 mu var0-0.534827
compute loss for weight  -0.399155  -0.399145 result 0.218526
 training batch 16 mu var0-0.534827
compute loss for weight  -0.39914  -0.399145 result 0.218524
 training batch 17 mu var0-0.534827
compute loss for weight  -0.39915  -0.399145 result 0.218525
   --dy = -0.131026 dy_ref = -0.131026
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |   0.002358      0.4347 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var0-0.534827
compute loss for weight  1.00001  1 result 0.218525
 training batch 19 mu var0-0.534827
compute loss for weight  0.99999  1 result 0.218525
 training batch 20 mu var0-0.534827
compute loss for weight  1.00001  1 result 0.218525
 training batch 21 mu var0-0.534827
compute loss for weight  0.999995  1 result 0.218525
   --dy = 0.00235802 dy_ref = 0.00235802
 training batch 22 mu var0-0.534827
compute loss for weight  1.00001  1 result 0.218529
 training batch 23 mu var0-0.534827
compute loss for weight  0.99999  1 result 0.21852
 training batch 24 mu var0-0.534827
compute loss for weight  1.00001  1 result 0.218527
 training batch 25 mu var0-0.534827
compute loss for weight  0.999995  1 result 0.218522
   --dy = 0.434691 dy_ref = 0.434691
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |  1.193e-17  -1.908e-17 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var0-0.534827
compute loss for weight  1e-05  0 result 0.218525
 training batch 27 mu var0-0.534827
compute loss for weight  -1e-05  0 result 0.218525
 training batch 28 mu var0-0.534827
compute loss for weight  5e-06  0 result 0.218525
 training batch 29 mu var0-0.534827
compute loss for weight  -5e-06  0 result 0.218525
   --dy = -3.70074e-12 dy_ref = 1.19262e-17
 training batch 30 mu var0-0.534827
compute loss for weight  1e-05  0 result 0.218525
 training batch 31 mu var0-0.534827
compute loss for weight  -1e-05  0 result 0.218525
 training batch 32 mu var0-0.534827
compute loss for weight  5e-06  0 result 0.218525
 training batch 33 mu var0-0.534827
compute loss for weight  -5e-06  0 result 0.218525
   --dy = -9.25186e-13 dy_ref = -1.9082e-17
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    0.01385     -0.8779 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.1702     -0.4952 

 training batch 34 mu var0-0.534827
compute loss for weight  0.170207  0.170197 result 0.218525
 training batch 35 mu var0-0.534827
compute loss for weight  0.170187  0.170197 result 0.218524
 training batch 36 mu var0-0.534827
compute loss for weight  0.170202  0.170197 result 0.218525
 training batch 37 mu var0-0.534827
compute loss for weight  0.170192  0.170197 result 0.218524
   --dy = 0.0138546 dy_ref = 0.0138546
 training batch 38 mu var0-0.534827
compute loss for weight  -0.495147  -0.495157 result 0.218516
 training batch 39 mu var0-0.534827
compute loss for weight  -0.495167  -0.495157 result 0.218533
 training batch 40 mu var0-0.534827
compute loss for weight  -0.495152  -0.495157 result 0.21852
 training batch 41 mu var0-0.534827
compute loss for weight  -0.495162  -0.495157 result 0.218529
   --dy = -0.877885 dy_ref = -0.877885
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m3.64722e-10[NON-XML-CHAR-0x1B][39m