Execution Time0.12s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: master-x86_64-centos7-gcc48-dbg (lcgapp-centos7-x86-64-25.cern.ch) on 2019-11-14 11:45:12

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 var00.145441
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.5251     -0.5556 
   1 |     -1.331     -0.3161 
   2 |      2.043     -0.9347 
   3 |    -0.9593      -1.308 
   4 |     0.4495     -0.7941 
   5 |      0.647      -1.493 
   6 |     -1.185       1.314 
   7 |      0.787      0.2749 
   8 |      0.781      0.2564 
   9 |    -0.3033    -0.06777 

output BN 
output DL feature 0 mean 0.145441	output DL std 1.06861
output DL feature 1 mean -0.362467	output DL std 0.844203
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.3745     -0.2411 
   1 |     -1.456     0.05786 
   2 |      1.872     -0.7145 
   3 |      -1.09      -1.181 
   4 |     0.2999     -0.5389 
   5 |     0.4947      -1.412 
   6 |     -1.313       2.093 
   7 |     0.6328      0.7957 
   8 |     0.6269      0.7726 
   9 |    -0.4426      0.3679 

output BN feature 0 mean -4.44089e-17	output BN std 1.05404
output BN feature 1 mean 1.55431e-16	output BN std 1.05401
Testing weight gradients   for    layer 0
weight gradient for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |    -0.3244      0.7459      0.1377      0.9551 
   1 |     0.7856      -0.535       1.538       -1.83 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |    -0.3231      0.2506       1.205     -0.4791 
   1 |    -0.3721      0.2504    -0.06722      0.7465 

 training batch 2 mu var00.145444
compute loss for weight  -0.323086  -0.323096 result 1.05274
 training batch 3 mu var00.145441
compute loss for weight  -0.323106  -0.323096 result 1.05275
 training batch 4 mu var00.145442
compute loss for weight  -0.323091  -0.323096 result 1.05274
 training batch 5 mu var00.145441
compute loss for weight  -0.323101  -0.323096 result 1.05275
   --dy = -0.324374 dy_ref = -0.324374
 training batch 6 mu var00.145441
compute loss for weight  0.250631  0.250621 result 1.05275
 training batch 7 mu var00.145441
compute loss for weight  0.250611  0.250621 result 1.05274
 training batch 8 mu var00.145441
compute loss for weight  0.250626  0.250621 result 1.05275
 training batch 9 mu var00.145441
compute loss for weight  0.250616  0.250621 result 1.05274
   --dy = 0.745895 dy_ref = 0.745895
 training batch 10 mu var00.145442
compute loss for weight  1.20544  1.20543 result 1.05274
 training batch 11 mu var00.145441
compute loss for weight  1.20542  1.20543 result 1.05274
 training batch 12 mu var00.145441
compute loss for weight  1.20543  1.20543 result 1.05274
 training batch 13 mu var00.145441
compute loss for weight  1.20542  1.20543 result 1.05274
   --dy = 0.137696 dy_ref = 0.137696
 training batch 14 mu var00.145441
compute loss for weight  -0.479093  -0.479103 result 1.05275
 training batch 15 mu var00.145441
compute loss for weight  -0.479113  -0.479103 result 1.05273
 training batch 16 mu var00.145441
compute loss for weight  -0.479098  -0.479103 result 1.05275
 training batch 17 mu var00.145441
compute loss for weight  -0.479108  -0.479103 result 1.05274
   --dy = 0.955071 dy_ref = 0.955071
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      1.498      0.6076 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var00.145441
compute loss for weight  1.00001  1 result 1.05276
 training batch 19 mu var00.145441
compute loss for weight  0.99999  1 result 1.05273
 training batch 20 mu var00.145441
compute loss for weight  1.00001  1 result 1.05275
 training batch 21 mu var00.145441
compute loss for weight  0.999995  1 result 1.05274
   --dy = 1.49784 dy_ref = 1.49784
 training batch 22 mu var00.145441
compute loss for weight  1.00001  1 result 1.05275
 training batch 23 mu var00.145441
compute loss for weight  0.99999  1 result 1.05274
 training batch 24 mu var00.145441
compute loss for weight  1.00001  1 result 1.05275
 training batch 25 mu var00.145441
compute loss for weight  0.999995  1 result 1.05274
   --dy = 0.607644 dy_ref = 0.607644
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |  1.804e-16   9.714e-17 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var00.145441
compute loss for weight  1e-05  0 result 1.05274
 training batch 27 mu var00.145441
compute loss for weight  -1e-05  0 result 1.05274
 training batch 28 mu var00.145441
compute loss for weight  5e-06  0 result 1.05274
 training batch 29 mu var00.145441
compute loss for weight  -5e-06  0 result 1.05274
   --dy = 2.96059e-11 dy_ref = 1.80411e-16
 training batch 30 mu var00.145441
compute loss for weight  1e-05  0 result 1.05274
 training batch 31 mu var00.145441
compute loss for weight  -1e-05  0 result 1.05274
 training batch 32 mu var00.145441
compute loss for weight  5e-06  0 result 1.05274
 training batch 33 mu var00.145441
compute loss for weight  -5e-06  0 result 1.05274
   --dy = 3.70074e-12 dy_ref = 9.71445e-17
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      1.528      0.8462 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.9802      0.7181 

 training batch 34 mu var00.145441
compute loss for weight  0.980169  0.980159 result 1.05276
 training batch 35 mu var00.145441
compute loss for weight  0.980149  0.980159 result 1.05273
 training batch 36 mu var00.145441
compute loss for weight  0.980164  0.980159 result 1.05275
 training batch 37 mu var00.145441
compute loss for weight  0.980154  0.980159 result 1.05274
   --dy = 1.52816 dy_ref = 1.52816
 training batch 38 mu var00.145441
compute loss for weight  0.718063  0.718053 result 1.05275
 training batch 39 mu var00.145441
compute loss for weight  0.718043  0.718053 result 1.05274
 training batch 40 mu var00.145441
compute loss for weight  0.718058  0.718053 result 1.05275
 training batch 41 mu var00.145441
compute loss for weight  0.718048  0.718053 result 1.05274
   --dy = 0.846238 dy_ref = 0.846238
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m4.35803e-10[NON-XML-CHAR-0x1B][39m