Execution Time0.08s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: master-x86_64-fedora29-gcc8 (root-fedora29-3.cern.ch) on 2019-11-13 04:39:59

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 var00.160869
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.1056      -1.902 
   1 |      0.686     -0.8515 
   2 |    -0.1264      -1.935 
   3 |      1.463      -4.087 
   4 |     0.3604      -2.614 
   5 |     0.6761      -4.935 
   6 |    -0.4234      0.9408 
   7 |    -0.6412       3.809 
   8 |    -0.6961     -0.2031 
   9 |     0.2047     -0.3579 

output BN 
output DL feature 0 mean 0.160869	output DL std 0.673569
output DL feature 1 mean -1.21354	output DL std 2.51257
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |   -0.08649     -0.2887 
   1 |     0.8217      0.1519 
   2 |    -0.4495     -0.3029 
   3 |      2.038      -1.205 
   4 |     0.3122     -0.5874 
   5 |     0.8062      -1.561 
   6 |    -0.9142      0.9038 
   7 |     -1.255       2.107 
   8 |     -1.341      0.4239 
   9 |    0.06862       0.359 

output BN feature 0 mean -8.60423e-17	output BN std 1.05396
output BN feature 1 mean 0	output BN std 1.05408
Testing weight gradients   for    layer 0
weight gradient for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |  -0.007665    0.007931    -0.01093  -0.0005085 
   1 |  -0.001461   0.0001732   -0.003427   -0.001525 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |     0.2882      -0.359     -0.4462     -0.3499 
   1 |    -0.9213       2.002     -0.1047      0.9748 

 training batch 2 mu var00.160872
compute loss for weight  0.288172  0.288162 result 0.00530614
 training batch 3 mu var00.160869
compute loss for weight  0.288152  0.288162 result 0.00530629
 training batch 4 mu var00.160869
compute loss for weight  0.288167  0.288162 result 0.00530618
 training batch 5 mu var00.160869
compute loss for weight  0.288157  0.288162 result 0.00530625
   --dy = -0.007665 dy_ref = -0.007665
 training batch 6 mu var00.160868
compute loss for weight  -0.35896  -0.35897 result 0.00530629
 training batch 7 mu var00.160869
compute loss for weight  -0.35898  -0.35897 result 0.00530614
 training batch 8 mu var00.160869
compute loss for weight  -0.358965  -0.35897 result 0.00530625
 training batch 9 mu var00.160869
compute loss for weight  -0.358975  -0.35897 result 0.00530618
   --dy = 0.00793053 dy_ref = 0.00793053
 training batch 10 mu var00.160869
compute loss for weight  -0.446199  -0.446209 result 0.00530611
 training batch 11 mu var00.160869
compute loss for weight  -0.446219  -0.446209 result 0.00530632
 training batch 12 mu var00.160869
compute loss for weight  -0.446204  -0.446209 result 0.00530616
 training batch 13 mu var00.160869
compute loss for weight  -0.446214  -0.446209 result 0.00530627
   --dy = -0.0109302 dy_ref = -0.0109302
 training batch 14 mu var00.160869
compute loss for weight  -0.349871  -0.349881 result 0.00530621
 training batch 15 mu var00.160869
compute loss for weight  -0.349891  -0.349881 result 0.00530622
 training batch 16 mu var00.160869
compute loss for weight  -0.349876  -0.349881 result 0.00530621
 training batch 17 mu var00.160869
compute loss for weight  -0.349886  -0.349881 result 0.00530622
   --dy = -0.000508487 dy_ref = -0.000508487
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |  -0.002107     0.01272 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var00.160869
compute loss for weight  1.00001  1 result 0.00530619
 training batch 19 mu var00.160869
compute loss for weight  0.99999  1 result 0.00530624
 training batch 20 mu var00.160869
compute loss for weight  1.00001  1 result 0.0053062
 training batch 21 mu var00.160869
compute loss for weight  0.999995  1 result 0.00530623
   --dy = -0.00210704 dy_ref = -0.00210704
 training batch 22 mu var00.160869
compute loss for weight  1.00001  1 result 0.00530634
 training batch 23 mu var00.160869
compute loss for weight  0.99999  1 result 0.00530609
 training batch 24 mu var00.160869
compute loss for weight  1.00001  1 result 0.00530628
 training batch 25 mu var00.160869
compute loss for weight  0.999995  1 result 0.00530615
   --dy = 0.0127195 dy_ref = 0.0127195
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 | -4.337e-19  -1.843e-18 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var00.160869
compute loss for weight  1e-05  0 result 0.00530622
 training batch 27 mu var00.160869
compute loss for weight  -1e-05  0 result 0.00530622
 training batch 28 mu var00.160869
compute loss for weight  5e-06  0 result 0.00530622
 training batch 29 mu var00.160869
compute loss for weight  -5e-06  0 result 0.00530622
   --dy = 1.15648e-13 dy_ref = -4.33681e-19
 training batch 30 mu var00.160869
compute loss for weight  1e-05  0 result 0.00530622
 training batch 31 mu var00.160869
compute loss for weight  -1e-05  0 result 0.00530622
 training batch 32 mu var00.160869
compute loss for weight  5e-06  0 result 0.00530622
 training batch 33 mu var00.160869
compute loss for weight  -5e-06  0 result 0.00530622
   --dy = 1.4456e-14 dy_ref = -1.84314e-18
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    0.03069     -0.1155 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |   -0.06866     -0.1101 

 training batch 34 mu var00.160869
compute loss for weight  -0.0686529  -0.0686629 result 0.00530652
 training batch 35 mu var00.160869
compute loss for weight  -0.0686729  -0.0686629 result 0.00530591
 training batch 36 mu var00.160869
compute loss for weight  -0.0686579  -0.0686629 result 0.00530637
 training batch 37 mu var00.160869
compute loss for weight  -0.0686679  -0.0686629 result 0.00530606
   --dy = 0.0306867 dy_ref = 0.0306867
 training batch 38 mu var00.160869
compute loss for weight  -0.110112  -0.110122 result 0.00530506
 training batch 39 mu var00.160869
compute loss for weight  -0.110132  -0.110122 result 0.00530737
 training batch 40 mu var00.160869
compute loss for weight  -0.110117  -0.110122 result 0.00530564
 training batch 41 mu var00.160869
compute loss for weight  -0.110127  -0.110122 result 0.00530679
   --dy = -0.115503 dy_ref = -0.115503
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m4.79184e-10[NON-XML-CHAR-0x1B][39m