rbm_out.log

Torch 7.0  Copyright (C) 2001-2011 Idiap, NEC Labs, NYU
<mnist> loading only 1000 examples	
<mnist> reading 1000 examples with 784+1 dimensions...	
<mnist> done	
<mnist> loading only 1000 examples	
<mnist> reading 1000 examples with 784+1 dimensions...	
<mnist> done	
1/10 - Recon err: 39.1	
2/10 - Recon err: 25.0	
3/10 - Recon err: 20.9	
4/10 - Recon err: 18.5	
5/10 - Recon err: 16.8	
6/10 - Recon err: 15.6	
7/10 - Recon err: 14.6	
8/10 - Recon err: 13.8	
9/10 - Recon err: 13.0	
10/10 - Recon err: 12.4	
{
  vW : FloatTensor - size: 7x3
  vb : FloatTensor - size: 3x1
  b : FloatTensor - size: 3x1
  vc : FloatTensor - size: 7x1
  dU : FloatTensor - size: 7x4
  U : FloatTensor - size: 7x4
  d : FloatTensor - size: 4x1
  n_classes : 4
  rand : function: 0x403806f0
  dW : FloatTensor - size: 7x3
  learningrate : 0.1
  W : FloatTensor - size: 7x3
  n_samples : 1
  db : FloatTensor - size: 3x1
  n_visible : 3
  dd : FloatTensor - size: 4x1
  dc : FloatTensor - size: 7x1
  vU : FloatTensor - size: 7x4
  c : FloatTensor - size: 7x1
  vd : FloatTensor - size: 4x1
  numepochs : 1
}
TEST of rbmup, sampler, rbmdownx, rbmdowny and samplevec : PASSED	
 0.4170
-0.2797
 0.0001
[torch.FloatTensor of dimension 3x1]

 0.4170
-0.2797
 0.0001
[torch.FloatTensor of dimension 3x1]

TEST of generative gradients                             : PASSED	
1/1 - Recon err:  0.3	
TEST of rbmtrain                                         : PASSED	
{
  vW : FloatTensor - size: 500x3
  vb : FloatTensor - size: 3x1
  b : FloatTensor - size: 3x1
  vc : FloatTensor - size: 500x1
  dU : FloatTensor - size: 500x4
  U : FloatTensor - size: 500x4
  d : FloatTensor - size: 4x1
  n_classes : 4
  rand : function: 0x412eec58
  dW : FloatTensor - size: 500x3
  learningrate : 0.05
  W : FloatTensor - size: 500x3
  n_samples : 1
  db : FloatTensor - size: 3x1
  n_visible : 3
  dd : FloatTensor - size: 4x1
  dc : FloatTensor - size: 500x1
  vU : FloatTensor - size: 500x4
  c : FloatTensor - size: 500x1
  vd : FloatTensor - size: 4x1
  numepochs : 10
}
1/10 - Recon err:  1.4	
2/10 - Recon err:  0.3	
3/10 - Recon err:  0.2	
4/10 - Recon err:  0.2	
5/10 - Recon err:  0.3	
6/10 - Recon err:  0.1	
7/10 - Recon err:  0.2	
8/10 - Recon err:  0.2	
9/10 - Recon err:  0.5	
10/10 - Recon err:  0.2