Execution Time0.11s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: master-x86_64-centos7-gcc48-dbg (lcgapp-centos7-x86-64-25.cern.ch) on 2019-11-13 14:36:51

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 var00.117484
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.4935      0.3753 
   1 |    -0.6933     -0.4241 
   2 |    0.01851      0.2795 
   3 |     -2.179    -0.07806 
   4 |  -0.009217      0.3699 
   5 |     0.3062      0.6565 
   6 |      1.537       1.106 
   7 |    -0.1963      -1.402 
   8 |      2.236      0.7805 
   9 |     -0.338   -0.008325 

output BN 
output DL feature 0 mean 0.117484	output DL std 1.19897
output DL feature 1 mean 0.165509	output DL std 0.706785
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.3306      0.3128 
   1 |    -0.7128     -0.8793 
   2 |   -0.08701        0.17 
   3 |     -2.019     -0.3632 
   4 |    -0.1114      0.3048 
   5 |     0.1659      0.7322 
   6 |      1.248       1.402 
   7 |    -0.2759      -2.337 
   8 |      1.862       0.917 
   9 |    -0.4004     -0.2592 

output BN feature 0 mean 9.4369e-17	output BN std 1.05405
output BN feature 1 mean -4.44089e-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 |     0.6128      -2.868      0.2368      0.1163 
   1 |     -10.79       25.41      -11.71      -13.53 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |    0.01628     0.09869      0.7011      0.9304 
   1 |    -0.1472      -0.625      0.3379       0.453 

 training batch 2 mu var00.117487
compute loss for weight  0.0162894  0.0162794 result 9.44122
 training batch 3 mu var00.117484
compute loss for weight  0.0162694  0.0162794 result 9.4412
 training batch 4 mu var00.117485
compute loss for weight  0.0162844  0.0162794 result 9.44121
 training batch 5 mu var00.117484
compute loss for weight  0.0162744  0.0162794 result 9.44121
   --dy = 0.612762 dy_ref = 0.612762
 training batch 6 mu var00.117484
compute loss for weight  0.0986983  0.0986883 result 9.44118
 training batch 7 mu var00.117484
compute loss for weight  0.0986783  0.0986883 result 9.44124
 training batch 8 mu var00.117484
compute loss for weight  0.0986933  0.0986883 result 9.4412
 training batch 9 mu var00.117484
compute loss for weight  0.0986833  0.0986883 result 9.44122
   --dy = -2.86826 dy_ref = -2.86826
 training batch 10 mu var00.117485
compute loss for weight  0.701145  0.701135 result 9.44121
 training batch 11 mu var00.117484
compute loss for weight  0.701125  0.701135 result 9.44121
 training batch 12 mu var00.117485
compute loss for weight  0.70114  0.701135 result 9.44121
 training batch 13 mu var00.117484
compute loss for weight  0.70113  0.701135 result 9.44121
   --dy = 0.236762 dy_ref = 0.236762
 training batch 14 mu var00.117484
compute loss for weight  0.930403  0.930393 result 9.44121
 training batch 15 mu var00.117484
compute loss for weight  0.930383  0.930393 result 9.44121
 training batch 16 mu var00.117484
compute loss for weight  0.930398  0.930393 result 9.44121
 training batch 17 mu var00.117484
compute loss for weight  0.930388  0.930393 result 9.44121
   --dy = 0.116292 dy_ref = 0.116292
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      14.37       4.508 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var00.117484
compute loss for weight  1.00001  1 result 9.44135
 training batch 19 mu var00.117484
compute loss for weight  0.99999  1 result 9.44107
 training batch 20 mu var00.117484
compute loss for weight  1.00001  1 result 9.44128
 training batch 21 mu var00.117484
compute loss for weight  0.999995  1 result 9.44114
   --dy = 14.3745 dy_ref = 14.3745
 training batch 22 mu var00.117484
compute loss for weight  1.00001  1 result 9.44125
 training batch 23 mu var00.117484
compute loss for weight  0.99999  1 result 9.44116
 training batch 24 mu var00.117484
compute loss for weight  1.00001  1 result 9.44123
 training batch 25 mu var00.117484
compute loss for weight  0.999995  1 result 9.44119
   --dy = 4.50789 dy_ref = 4.50789
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |  1.332e-15   3.886e-16 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var00.117484
compute loss for weight  1e-05  0 result 9.44121
 training batch 27 mu var00.117484
compute loss for weight  -1e-05  0 result 9.44121
 training batch 28 mu var00.117484
compute loss for weight  5e-06  0 result 9.44121
 training batch 29 mu var00.117484
compute loss for weight  -5e-06  0 result 9.44121
   --dy = 2.36848e-10 dy_ref = 1.33227e-15
 training batch 30 mu var00.117484
compute loss for weight  1e-05  0 result 9.44121
 training batch 31 mu var00.117484
compute loss for weight  -1e-05  0 result 9.44121
 training batch 32 mu var00.117484
compute loss for weight  5e-06  0 result 9.44121
 training batch 33 mu var00.117484
compute loss for weight  -5e-06  0 result 9.44121
   --dy = 2.96059e-11 dy_ref = 3.88578e-16
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      5.941       4.697 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |       2.42      0.9597 

 training batch 34 mu var00.117484
compute loss for weight  2.41963  2.41962 result 9.44127
 training batch 35 mu var00.117484
compute loss for weight  2.41961  2.41962 result 9.44115
 training batch 36 mu var00.117484
compute loss for weight  2.41962  2.41962 result 9.44124
 training batch 37 mu var00.117484
compute loss for weight  2.41961  2.41962 result 9.44118
   --dy = 5.94083 dy_ref = 5.94083
 training batch 38 mu var00.117484
compute loss for weight  0.959681  0.959671 result 9.44126
 training batch 39 mu var00.117484
compute loss for weight  0.959661  0.959671 result 9.44116
 training batch 40 mu var00.117484
compute loss for weight  0.959676  0.959671 result 9.44123
 training batch 41 mu var00.117484
compute loss for weight  0.959666  0.959671 result 9.44119
   --dy = 4.69733 dy_ref = 4.69733
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][33m1.21431e-09[NON-XML-CHAR-0x1B][39m