Execution Time0.97s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: PR-4623-x86_64-centos7-gcc48-opt (olhswep22.cern.ch) on 2019-11-14 10:10:23

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 var00.849833
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      1.615     -0.2173 
   1 |      0.375     -0.6034 
   2 |      2.059      0.4843 
   3 |   -0.06684      0.7949 
   4 |      1.419     0.08771 
   5 |      3.257     -0.2983 
   6 |       -2.9      0.3282 
   7 |      1.464      -1.087 
   8 |      1.652     -0.7886 
   9 |    -0.3753      0.1646 

output BN 
output DL feature 0 mean 0.849833	output DL std 1.69378
output DL feature 1 mean -0.113512	output DL std 0.594793
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      0.476     -0.1838 
   1 |    -0.2955      -0.868 
   2 |     0.7525       1.059 
   3 |    -0.5705        1.61 
   4 |     0.3541      0.3566 
   5 |      1.498     -0.3275 
   6 |     -2.334      0.7826 
   7 |     0.3823      -1.725 
   8 |      0.499      -1.196 
   9 |    -0.7624      0.4928 

output BN feature 0 mean 4.44089e-17	output BN std 1.05407
output BN feature 1 mean 5.55112e-18	output BN std 1.05393
Testing weight gradients   for    layer 0
weight gradient for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |  -0.005571    -0.03011    0.001151    -0.01996 
   1 |     0.1964       0.298      0.1606     0.05337 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |      1.238      0.6205      0.7805      -1.237 
   1 |    -0.6092     -0.6232      0.1532    -0.02026 

 training batch 2 mu var00.849836
compute loss for weight  1.2385  1.23849 result 0.0576172
 training batch 3 mu var00.849833
compute loss for weight  1.23848  1.23849 result 0.0576173
 training batch 4 mu var00.849834
compute loss for weight  1.23849  1.23849 result 0.0576172
 training batch 5 mu var00.849833
compute loss for weight  1.23848  1.23849 result 0.0576173
   --dy = -0.00557087 dy_ref = -0.00557087
 training batch 6 mu var00.849833
compute loss for weight  0.620492  0.620482 result 0.0576169
 training batch 7 mu var00.849833
compute loss for weight  0.620472  0.620482 result 0.0576175
 training batch 8 mu var00.849833
compute loss for weight  0.620487  0.620482 result 0.0576171
 training batch 9 mu var00.849833
compute loss for weight  0.620477  0.620482 result 0.0576174
   --dy = -0.0301106 dy_ref = -0.0301106
 training batch 10 mu var00.849834
compute loss for weight  0.780464  0.780454 result 0.0576173
 training batch 11 mu var00.849833
compute loss for weight  0.780444  0.780454 result 0.0576172
 training batch 12 mu var00.849834
compute loss for weight  0.780459  0.780454 result 0.0576172
 training batch 13 mu var00.849833
compute loss for weight  0.780449  0.780454 result 0.0576172
   --dy = 0.00115108 dy_ref = 0.00115108
 training batch 14 mu var00.849833
compute loss for weight  -1.23686  -1.23687 result 0.057617
 training batch 15 mu var00.849833
compute loss for weight  -1.23688  -1.23687 result 0.0576174
 training batch 16 mu var00.849833
compute loss for weight  -1.23686  -1.23687 result 0.0576171
 training batch 17 mu var00.849833
compute loss for weight  -1.23687  -1.23687 result 0.0576173
   --dy = -0.0199601 dy_ref = -0.0199601
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    0.09904      0.0162 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var00.849833
compute loss for weight  1.00001  1 result 0.0576182
 training batch 19 mu var00.849833
compute loss for weight  0.99999  1 result 0.0576163
 training batch 20 mu var00.849833
compute loss for weight  1.00001  1 result 0.0576177
 training batch 21 mu var00.849833
compute loss for weight  0.999995  1 result 0.0576167
   --dy = 0.0990351 dy_ref = 0.0990351
 training batch 22 mu var00.849833
compute loss for weight  1.00001  1 result 0.0576174
 training batch 23 mu var00.849833
compute loss for weight  0.99999  1 result 0.0576171
 training batch 24 mu var00.849833
compute loss for weight  1.00001  1 result 0.0576173
 training batch 25 mu var00.849833
compute loss for weight  0.999995  1 result 0.0576172
   --dy = 0.0161994 dy_ref = 0.0161994
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |  1.041e-17   6.072e-18 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var00.849833
compute loss for weight  1e-05  0 result 0.0576172
 training batch 27 mu var00.849833
compute loss for weight  -1e-05  0 result 0.0576172
 training batch 28 mu var00.849833
compute loss for weight  5e-06  0 result 0.0576172
 training batch 29 mu var00.849833
compute loss for weight  -5e-06  0 result 0.0576172
   --dy = 2.19732e-12 dy_ref = 1.04083e-17
 training batch 30 mu var00.849833
compute loss for weight  1e-05  0 result 0.0576172
 training batch 31 mu var00.849833
compute loss for weight  -1e-05  0 result 0.0576172
 training batch 32 mu var00.849833
compute loss for weight  5e-06  0 result 0.0576172
 training batch 33 mu var00.849833
compute loss for weight  -5e-06  0 result 0.0576172
   --dy = -3.46945e-13 dy_ref = 6.07153e-18
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.3922      0.1083 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.2525      0.1495 

 training batch 34 mu var00.849833
compute loss for weight  0.252553  0.252543 result 0.0576212
 training batch 35 mu var00.849833
compute loss for weight  0.252533  0.252543 result 0.0576133
 training batch 36 mu var00.849833
compute loss for weight  0.252548  0.252543 result 0.0576192
 training batch 37 mu var00.849833
compute loss for weight  0.252538  0.252543 result 0.0576153
   --dy = 0.392151 dy_ref = 0.392151
 training batch 38 mu var00.849833
compute loss for weight  0.149559  0.149549 result 0.0576183
 training batch 39 mu var00.849833
compute loss for weight  0.149539  0.149549 result 0.0576162
 training batch 40 mu var00.849833
compute loss for weight  0.149554  0.149549 result 0.0576178
 training batch 41 mu var00.849833
compute loss for weight  0.149544  0.149549 result 0.0576167
   --dy = 0.108322 dy_ref = 0.108322
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][33m1.00526e-09[NON-XML-CHAR-0x1B][39m