Execution Time0.06s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: master-x86_64-fedora30-gcc9 (root-fedora30-1.cern.ch) on 2019-11-13 01:05:01

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 var0-0.0118856
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    0.02913     -0.9888 
   1 |      1.126       -1.03 
   2 |     -2.289      0.4062 
   3 |     -2.212     -0.5648 
   4 |    -0.7114     -0.8698 
   5 |    -0.4078      -2.214 
   6 |      1.707     -0.6964 
   7 |     0.9404       1.385 
   8 |      1.947      -1.411 
   9 |    -0.2483   -0.009899 

output BN 
output DL feature 0 mean -0.0118856	output DL std 1.48029
output DL feature 1 mean -0.59928	output DL std 0.99897
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.0292      -0.411 
   1 |       0.81     -0.4545 
   2 |     -1.621       1.061 
   3 |     -1.567     0.03634 
   4 |    -0.4981     -0.2855 
   5 |    -0.2819      -1.703 
   6 |      1.224     -0.1025 
   7 |     0.6781       2.094 
   8 |      1.395      -0.856 
   9 |    -0.1684      0.6219 

output BN feature 0 mean -8.32667e-18	output BN std 1.05407
output BN feature 1 mean -6.66134e-17	output BN std 1.05403
Testing weight gradients   for    layer 0
weight gradient for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |    0.03896    -0.04272     0.01316    0.007389 
   1 |      0.109     -0.1597     0.04011    -0.02667 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |     0.8834       0.824     -0.5889       1.155 
   1 |    -0.9685      0.5151      0.2483     -0.2843 

 training batch 2 mu var0-0.0118828
compute loss for weight  0.883447  0.883437 result 0.103336
 training batch 3 mu var0-0.0118856
compute loss for weight  0.883427  0.883437 result 0.103335
 training batch 4 mu var0-0.0118849
compute loss for weight  0.883442  0.883437 result 0.103336
 training batch 5 mu var0-0.0118856
compute loss for weight  0.883432  0.883437 result 0.103336
   --dy = 0.0389578 dy_ref = 0.0389578
 training batch 6 mu var0-0.0118861
compute loss for weight  0.824038  0.824028 result 0.103335
 training batch 7 mu var0-0.0118856
compute loss for weight  0.824018  0.824028 result 0.103336
 training batch 8 mu var0-0.0118858
compute loss for weight  0.824033  0.824028 result 0.103336
 training batch 9 mu var0-0.0118856
compute loss for weight  0.824023  0.824028 result 0.103336
   --dy = -0.042716 dy_ref = -0.042716
 training batch 10 mu var0-0.0118853
compute loss for weight  -0.588877  -0.588887 result 0.103336
 training batch 11 mu var0-0.0118856
compute loss for weight  -0.588897  -0.588887 result 0.103336
 training batch 12 mu var0-0.0118855
compute loss for weight  -0.588882  -0.588887 result 0.103336
 training batch 13 mu var0-0.0118856
compute loss for weight  -0.588892  -0.588887 result 0.103336
   --dy = 0.0131564 dy_ref = 0.0131564
 training batch 14 mu var0-0.0118857
compute loss for weight  1.15531  1.1553 result 0.103336
 training batch 15 mu var0-0.0118856
compute loss for weight  1.15529  1.1553 result 0.103336
 training batch 16 mu var0-0.0118857
compute loss for weight  1.1553  1.1553 result 0.103336
 training batch 17 mu var0-0.0118856
compute loss for weight  1.15529  1.1553 result 0.103336
   --dy = 0.00738856 dy_ref = 0.00738856
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      0.115      0.0917 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var0-0.0118856
compute loss for weight  1.00001  1 result 0.103337
 training batch 19 mu var0-0.0118856
compute loss for weight  0.99999  1 result 0.103335
 training batch 20 mu var0-0.0118856
compute loss for weight  1.00001  1 result 0.103336
 training batch 21 mu var0-0.0118856
compute loss for weight  0.999995  1 result 0.103335
   --dy = 0.114972 dy_ref = 0.114972
 training batch 22 mu var0-0.0118856
compute loss for weight  1.00001  1 result 0.103337
 training batch 23 mu var0-0.0118856
compute loss for weight  0.99999  1 result 0.103335
 training batch 24 mu var0-0.0118856
compute loss for weight  1.00001  1 result 0.103336
 training batch 25 mu var0-0.0118856
compute loss for weight  0.999995  1 result 0.103335
   --dy = 0.0916999 dy_ref = 0.0916999
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |  7.806e-18  -7.806e-18 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var0-0.0118856
compute loss for weight  1e-05  0 result 0.103336
 training batch 27 mu var0-0.0118856
compute loss for weight  -1e-05  0 result 0.103336
 training batch 28 mu var0-0.0118856
compute loss for weight  5e-06  0 result 0.103336
 training batch 29 mu var0-0.0118856
compute loss for weight  -5e-06  0 result 0.103336
   --dy = 2.31296e-12 dy_ref = 7.80626e-18
 training batch 30 mu var0-0.0118856
compute loss for weight  1e-05  0 result 0.103336
 training batch 31 mu var0-0.0118856
compute loss for weight  -1e-05  0 result 0.103336
 training batch 32 mu var0-0.0118856
compute loss for weight  5e-06  0 result 0.103336
 training batch 33 mu var0-0.0118856
compute loss for weight  -5e-06  0 result 0.103336
   --dy = -4.62593e-13 dy_ref = -7.80626e-18
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.5109      0.4643 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     -0.225      0.1975 

 training batch 34 mu var0-0.0118856
compute loss for weight  -0.225039  -0.225049 result 0.103331
 training batch 35 mu var0-0.0118856
compute loss for weight  -0.225059  -0.225049 result 0.103341
 training batch 36 mu var0-0.0118856
compute loss for weight  -0.225044  -0.225049 result 0.103333
 training batch 37 mu var0-0.0118856
compute loss for weight  -0.225054  -0.225049 result 0.103338
   --dy = -0.510874 dy_ref = -0.510874
 training batch 38 mu var0-0.0118856
compute loss for weight  0.197527  0.197517 result 0.10334
 training batch 39 mu var0-0.0118856
compute loss for weight  0.197507  0.197517 result 0.103331
 training batch 40 mu var0-0.0118856
compute loss for weight  0.197522  0.197517 result 0.103338
 training batch 41 mu var0-0.0118856
compute loss for weight  0.197512  0.197517 result 0.103334
   --dy = 0.464263 dy_ref = 0.464263
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m1.79347e-10[NON-XML-CHAR-0x1B][39m