Execution Time0.11s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: master-x86_64-centos7-gcc48 (olhswep22.cern.ch) on 2019-11-15 01:05:03

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 var00.337392
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.4417     -0.3658 
   1 |     0.6245     -0.2156 
   2 |     0.9915       1.372 
   3 |      1.667       1.681 
   4 |     0.7457      0.1513 
   5 |      1.411     -0.2747 
   6 |     -2.283      -2.715 
   7 |     0.4997       1.105 
   8 |    -0.8172      -2.212 
   9 |    0.09195      0.1666 

output BN 
output DL feature 0 mean 0.337392	output DL std 1.14963
output DL feature 1 mean -0.130773	output DL std 1.42713
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    0.09565     -0.1736 
   1 |     0.2633    -0.06263 
   2 |     0.5997        1.11 
   3 |      1.219       1.338 
   4 |     0.3743      0.2083 
   5 |     0.9847     -0.1063 
   6 |     -2.402      -1.909 
   7 |     0.1488      0.9128 
   8 |     -1.059      -1.537 
   9 |     -0.225      0.2196 

output BN feature 0 mean -7.77156e-17	output BN std 1.05405
output BN feature 1 mean 4.16334e-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 |     0.4518     -0.2463      0.2032       0.166 
   1 |    -0.3639      0.2264     -0.1453     0.04591 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |     0.5252     0.06191     -0.1176      -1.194 
   1 |    -0.3436      0.2075     0.03115      -1.438 

 training batch 2 mu var00.337395
compute loss for weight  0.525226  0.525216 result 0.241389
 training batch 3 mu var00.337392
compute loss for weight  0.525206  0.525216 result 0.24138
 training batch 4 mu var00.337393
compute loss for weight  0.525221  0.525216 result 0.241386
 training batch 5 mu var00.337392
compute loss for weight  0.525211  0.525216 result 0.241382
   --dy = 0.451841 dy_ref = 0.451841
 training batch 6 mu var00.337392
compute loss for weight  0.061918  0.061908 result 0.241382
 training batch 7 mu var00.337392
compute loss for weight  0.061898  0.061908 result 0.241387
 training batch 8 mu var00.337392
compute loss for weight  0.061913  0.061908 result 0.241383
 training batch 9 mu var00.337392
compute loss for weight  0.061903  0.061908 result 0.241385
   --dy = -0.246278 dy_ref = -0.246278
 training batch 10 mu var00.337392
compute loss for weight  -0.117549  -0.117559 result 0.241386
 training batch 11 mu var00.337392
compute loss for weight  -0.117569  -0.117559 result 0.241382
 training batch 12 mu var00.337392
compute loss for weight  -0.117554  -0.117559 result 0.241385
 training batch 13 mu var00.337392
compute loss for weight  -0.117564  -0.117559 result 0.241383
   --dy = 0.20324 dy_ref = 0.20324
 training batch 14 mu var00.337392
compute loss for weight  -1.19414  -1.19415 result 0.241386
 training batch 15 mu var00.337392
compute loss for weight  -1.19416  -1.19415 result 0.241383
 training batch 16 mu var00.337392
compute loss for weight  -1.19414  -1.19415 result 0.241385
 training batch 17 mu var00.337392
compute loss for weight  -1.19415  -1.19415 result 0.241383
   --dy = 0.165964 dy_ref = 0.165964
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.1259      0.6087 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var00.337392
compute loss for weight  1.00001  1 result 0.241383
 training batch 19 mu var00.337392
compute loss for weight  0.99999  1 result 0.241385
 training batch 20 mu var00.337392
compute loss for weight  1.00001  1 result 0.241384
 training batch 21 mu var00.337392
compute loss for weight  0.999995  1 result 0.241385
   --dy = -0.12592 dy_ref = -0.12592
 training batch 22 mu var00.337392
compute loss for weight  1.00001  1 result 0.24139
 training batch 23 mu var00.337392
compute loss for weight  0.99999  1 result 0.241378
 training batch 24 mu var00.337392
compute loss for weight  1.00001  1 result 0.241387
 training batch 25 mu var00.337392
compute loss for weight  0.999995  1 result 0.241381
   --dy = 0.608688 dy_ref = 0.608688
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 | -9.714e-17   1.388e-16 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var00.337392
compute loss for weight  1e-05  0 result 0.241384
 training batch 27 mu var00.337392
compute loss for weight  -1e-05  0 result 0.241384
 training batch 28 mu var00.337392
compute loss for weight  5e-06  0 result 0.241384
 training batch 29 mu var00.337392
compute loss for weight  -5e-06  0 result 0.241384
   --dy = 0 dy_ref = -9.71445e-17
 training batch 30 mu var00.337392
compute loss for weight  1e-05  0 result 0.241384
 training batch 31 mu var00.337392
compute loss for weight  -1e-05  0 result 0.241384
 training batch 32 mu var00.337392
compute loss for weight  5e-06  0 result 0.241384
 training batch 33 mu var00.337392
compute loss for weight  -5e-06  0 result 0.241384
   --dy = 4.16334e-12 dy_ref = 1.38778e-16
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.1792      0.6561 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.7025      0.9278 

 training batch 34 mu var00.337392
compute loss for weight  -0.702494  -0.702504 result 0.241386
 training batch 35 mu var00.337392
compute loss for weight  -0.702514  -0.702504 result 0.241382
 training batch 36 mu var00.337392
compute loss for weight  -0.702499  -0.702504 result 0.241385
 training batch 37 mu var00.337392
compute loss for weight  -0.702509  -0.702504 result 0.241383
   --dy = 0.179245 dy_ref = 0.179245
 training batch 38 mu var00.337392
compute loss for weight  0.927815  0.927805 result 0.241391
 training batch 39 mu var00.337392
compute loss for weight  0.927795  0.927805 result 0.241378
 training batch 40 mu var00.337392
compute loss for weight  0.92781  0.927805 result 0.241387
 training batch 41 mu var00.337392
compute loss for weight  0.9278  0.927805 result 0.241381
   --dy = 0.656052 dy_ref = 0.656052
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m2.90627e-11[NON-XML-CHAR-0x1B][39m