Execution Time0.14s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: master-x86_64-fedora29-gcc8-dbg (root-fedora29-2.cern.ch) on 2019-11-14 11:43:11

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 var00.796165
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      1.585     -0.1557 
   1 |     0.5586     -0.4639 
   2 |     -0.511      0.4193 
   3 |     -1.326        1.33 
   4 |     0.9899      0.2525 
   5 |      2.822    -0.02401 
   6 |       2.34      0.5533 
   7 |     -1.896      -1.725 
   8 |      3.687     -0.9346 
   9 |    -0.2873      0.2554 

output BN 
output DL feature 0 mean 0.796165	output DL std 1.8317
output DL feature 1 mean -0.0492301	output DL std 0.846214
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.4539     -0.1327 
   1 |    -0.1367     -0.5165 
   2 |    -0.7522      0.5836 
   3 |     -1.221       1.718 
   4 |     0.1115      0.3758 
   5 |      1.166     0.03142 
   6 |     0.8884      0.7505 
   7 |     -1.549      -2.087 
   8 |      1.663      -1.103 
   9 |    -0.6235      0.3794 

output BN feature 0 mean -5.55112e-17	output BN std 1.05408
output BN feature 1 mean -4.996e-17	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 |    -0.1284     -0.7929     -0.2063     -0.1828 
   1 |      1.948        2.58       2.006       2.287 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |      1.119     -0.5533      0.3367       1.234 
   1 |     -0.589     -0.9339     0.03425    0.007376 

 training batch 2 mu var00.796168
compute loss for weight  1.11946  1.11945 result 1.64687
 training batch 3 mu var00.796165
compute loss for weight  1.11944  1.11945 result 1.64688
 training batch 4 mu var00.796166
compute loss for weight  1.11945  1.11945 result 1.64688
 training batch 5 mu var00.796165
compute loss for weight  1.11944  1.11945 result 1.64688
   --dy = -0.128354 dy_ref = -0.128354
 training batch 6 mu var00.796165
compute loss for weight  -0.553275  -0.553285 result 1.64687
 training batch 7 mu var00.796165
compute loss for weight  -0.553295  -0.553285 result 1.64688
 training batch 8 mu var00.796165
compute loss for weight  -0.55328  -0.553285 result 1.64687
 training batch 9 mu var00.796165
compute loss for weight  -0.55329  -0.553285 result 1.64688
   --dy = -0.792947 dy_ref = -0.792947
 training batch 10 mu var00.796165
compute loss for weight  0.336681  0.336671 result 1.64687
 training batch 11 mu var00.796165
compute loss for weight  0.336661  0.336671 result 1.64688
 training batch 12 mu var00.796165
compute loss for weight  0.336676  0.336671 result 1.64687
 training batch 13 mu var00.796165
compute loss for weight  0.336666  0.336671 result 1.64688
   --dy = -0.206345 dy_ref = -0.206345
 training batch 14 mu var00.796165
compute loss for weight  1.23398  1.23397 result 1.64687
 training batch 15 mu var00.796165
compute loss for weight  1.23396  1.23397 result 1.64688
 training batch 16 mu var00.796165
compute loss for weight  1.23398  1.23397 result 1.64687
 training batch 17 mu var00.796165
compute loss for weight  1.23397  1.23397 result 1.64688
   --dy = -0.182758 dy_ref = -0.182758
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      1.604       1.689 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var00.796165
compute loss for weight  1.00001  1 result 1.64689
 training batch 19 mu var00.796165
compute loss for weight  0.99999  1 result 1.64686
 training batch 20 mu var00.796165
compute loss for weight  1.00001  1 result 1.64688
 training batch 21 mu var00.796165
compute loss for weight  0.999995  1 result 1.64687
   --dy = 1.60428 dy_ref = 1.60428
 training batch 22 mu var00.796165
compute loss for weight  1.00001  1 result 1.64689
 training batch 23 mu var00.796165
compute loss for weight  0.99999  1 result 1.64686
 training batch 24 mu var00.796165
compute loss for weight  1.00001  1 result 1.64688
 training batch 25 mu var00.796165
compute loss for weight  0.999995  1 result 1.64687
   --dy = 1.68947 dy_ref = 1.68947
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 | -9.021e-17  -1.041e-16 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var00.796165
compute loss for weight  1e-05  0 result 1.64688
 training batch 27 mu var00.796165
compute loss for weight  -1e-05  0 result 1.64688
 training batch 28 mu var00.796165
compute loss for weight  5e-06  0 result 1.64688
 training batch 29 mu var00.796165
compute loss for weight  -5e-06  0 result 1.64688
   --dy = -7.40149e-12 dy_ref = -9.02056e-17
 training batch 30 mu var00.796165
compute loss for weight  1e-05  0 result 1.64688
 training batch 31 mu var00.796165
compute loss for weight  -1e-05  0 result 1.64688
 training batch 32 mu var00.796165
compute loss for weight  5e-06  0 result 1.64688
 training batch 33 mu var00.796165
compute loss for weight  -5e-06  0 result 1.64688
   --dy = 8.88178e-11 dy_ref = -1.04083e-16
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |      1.733       1.782 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.9255      0.9483 

 training batch 34 mu var00.796165
compute loss for weight  0.925513  0.925503 result 1.64689
 training batch 35 mu var00.796165
compute loss for weight  0.925493  0.925503 result 1.64686
 training batch 36 mu var00.796165
compute loss for weight  0.925508  0.925503 result 1.64688
 training batch 37 mu var00.796165
compute loss for weight  0.925498  0.925503 result 1.64687
   --dy = 1.73342 dy_ref = 1.73342
 training batch 38 mu var00.796165
compute loss for weight  0.948304  0.948294 result 1.64689
 training batch 39 mu var00.796165
compute loss for weight  0.948284  0.948294 result 1.64686
 training batch 40 mu var00.796165
compute loss for weight  0.948299  0.948294 result 1.64688
 training batch 41 mu var00.796165
compute loss for weight  0.948289  0.948294 result 1.64687
   --dy = 1.78159 dy_ref = 1.78159
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m7.13699e-10[NON-XML-CHAR-0x1B][39m