Execution Time0.17s

Test: TMVA-DNN-BatchNormalization (Passed)
Build: master-x86_64-centos7-clang100-opt-master (olsnba08.cern.ch) on 2019-11-12 23:51:49
Repository revision: 30660dce2d9e89e4852dbf83dbd8b2cfcc137eff

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 var0-0.0750332
output DL 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.2224       -1.73 
   1 |     0.1029      -0.999 
   2 |      -0.32     -0.0742 
   3 |      0.741      -0.949 
   4 |   -0.03821      -1.632 
   5 |    -0.2413      -3.768 
   6 |     0.4974      -1.286 
   7 |    -0.8408       2.838 
   8 |    -0.6042      -2.374 
   9 |     0.1752    -0.00395 

output BN 
output DL feature 0 mean -0.0750332	output DL std 0.479015
output DL feature 1 mean -0.997755	output DL std 1.73677
output of BN 

10x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.3242     -0.4446 
   1 |     0.3915  -0.0007504 
   2 |     -0.539      0.5605 
   3 |      1.795     0.02961 
   4 |    0.08102      -0.385 
   5 |    -0.3658      -1.681 
   6 |      1.259     -0.1748 
   7 |     -1.685       2.328 
   8 |     -1.164     -0.8352 
   9 |     0.5506      0.6032 

output BN feature 0 mean -1.11022e-17	output BN std 1.05384
output BN feature 1 mean 0	output BN std 1.05407
Testing weight gradients   for    layer 0
weight gradient for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |     0.8773     -0.7072      0.5985      0.2798 
   1 |        0.2      0.0944      0.2693     0.04862 

weights for layer 0

2x4 matrix is as follows

     |      0    |      1    |      2    |      3    |
---------------------------------------------------------
   0 |    -0.1895      -0.419     -0.2709      0.1155 
   1 |     -1.209       1.175    -0.03263     -0.4704 

 training batch 2 mu var0-0.0750304
compute loss for weight  -0.189534  -0.189544 result 0.761796
 training batch 3 mu var0-0.0750332
compute loss for weight  -0.189554  -0.189544 result 0.761778
 training batch 4 mu var0-0.0750325
compute loss for weight  -0.189539  -0.189544 result 0.761792
 training batch 5 mu var0-0.0750332
compute loss for weight  -0.189549  -0.189544 result 0.761783
   --dy = 0.877301 dy_ref = 0.877301
 training batch 6 mu var0-0.0750337
compute loss for weight  -0.419037  -0.419047 result 0.76178
 training batch 7 mu var0-0.0750332
compute loss for weight  -0.419057  -0.419047 result 0.761794
 training batch 8 mu var0-0.0750334
compute loss for weight  -0.419042  -0.419047 result 0.761784
 training batch 9 mu var0-0.0750332
compute loss for weight  -0.419052  -0.419047 result 0.761791
   --dy = -0.707247 dy_ref = -0.707247
 training batch 10 mu var0-0.0750329
compute loss for weight  -0.270873  -0.270883 result 0.761793
 training batch 11 mu var0-0.0750332
compute loss for weight  -0.270893  -0.270883 result 0.761781
 training batch 12 mu var0-0.0750331
compute loss for weight  -0.270878  -0.270883 result 0.76179
 training batch 13 mu var0-0.0750332
compute loss for weight  -0.270888  -0.270883 result 0.761784
   --dy = 0.598503 dy_ref = 0.598503
 training batch 14 mu var0-0.0750333
compute loss for weight  0.11553  0.11552 result 0.76179
 training batch 15 mu var0-0.0750332
compute loss for weight  0.11551  0.11552 result 0.761784
 training batch 16 mu var0-0.0750332
compute loss for weight  0.115525  0.11552 result 0.761789
 training batch 17 mu var0-0.0750332
compute loss for weight  0.115515  0.11552 result 0.761786
   --dy = 0.279839 dy_ref = 0.279839
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     0.5898      0.9337 

weights for layer 1

1x2 matrix is as follows

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

 training batch 18 mu var0-0.0750332
compute loss for weight  1.00001  1 result 0.761793
 training batch 19 mu var0-0.0750332
compute loss for weight  0.99999  1 result 0.761781
 training batch 20 mu var0-0.0750332
compute loss for weight  1.00001  1 result 0.76179
 training batch 21 mu var0-0.0750332
compute loss for weight  0.999995  1 result 0.761784
   --dy = 0.589846 dy_ref = 0.589846
 training batch 22 mu var0-0.0750332
compute loss for weight  1.00001  1 result 0.761797
 training batch 23 mu var0-0.0750332
compute loss for weight  0.99999  1 result 0.761778
 training batch 24 mu var0-0.0750332
compute loss for weight  1.00001  1 result 0.761792
 training batch 25 mu var0-0.0750332
compute loss for weight  0.999995  1 result 0.761783
   --dy = 0.933729 dy_ref = 0.933729
Testing weight gradients   for    layer 1
weight gradient for layer 1

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |  2.255e-17  -2.082e-17 

weights for layer 1

1x2 matrix is as follows

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

 training batch 26 mu var0-0.0750332
compute loss for weight  1e-05  0 result 0.761787
 training batch 27 mu var0-0.0750332
compute loss for weight  -1e-05  0 result 0.761787
 training batch 28 mu var0-0.0750332
compute loss for weight  5e-06  0 result 0.761787
 training batch 29 mu var0-0.0750332
compute loss for weight  -5e-06  0 result 0.761787
   --dy = -2.96059e-11 dy_ref = 2.25514e-17
 training batch 30 mu var0-0.0750332
compute loss for weight  1e-05  0 result 0.761787
 training batch 31 mu var0-0.0750332
compute loss for weight  -1e-05  0 result 0.761787
 training batch 32 mu var0-0.0750332
compute loss for weight  5e-06  0 result 0.761787
 training batch 33 mu var0-0.0750332
compute loss for weight  -5e-06  0 result 0.761787
   --dy = -1.85037e-12 dy_ref = -2.08167e-17
Testing weight gradients   for    layer 2
weight gradient for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |     -1.245       1.483 

weights for layer 2

1x2 matrix is as follows

     |      0    |      1    |
-------------------------------
   0 |    -0.4739      0.6297 

 training batch 34 mu var0-0.0750332
compute loss for weight  -0.473934  -0.473944 result 0.761775
 training batch 35 mu var0-0.0750332
compute loss for weight  -0.473954  -0.473944 result 0.7618
 training batch 36 mu var0-0.0750332
compute loss for weight  -0.473939  -0.473944 result 0.761781
 training batch 37 mu var0-0.0750332
compute loss for weight  -0.473949  -0.473944 result 0.761793
   --dy = -1.24455 dy_ref = -1.24455
 training batch 38 mu var0-0.0750332
compute loss for weight  0.629669  0.629659 result 0.761802
 training batch 39 mu var0-0.0750332
compute loss for weight  0.629649  0.629659 result 0.761772
 training batch 40 mu var0-0.0750332
compute loss for weight  0.629664  0.629659 result 0.761795
 training batch 41 mu var0-0.0750332
compute loss for weight  0.629654  0.629659 result 0.76178
   --dy = 1.48291 dy_ref = 1.48291
Testing weight gradients:      maximum relative error: [NON-XML-CHAR-0x1B][32m8.29372e-11[NON-XML-CHAR-0x1B][39m