Commits · 351a6331e9501777387f7e96cae8948912fc9f9c · OpenDAS / dlib

"vscode:/vscode.git/clone" did not exist on "e50b89be1b948faffc4dbdf124af7411a9e87ff0"

13 Dec, 2015 2 commits
- Added loss_multiclass_log_ · 351a6331
  Davis King authored Dec 13, 2015
  
  351a6331
- Fixed errant const. · 04526626
  Davis King authored Dec 13, 2015
  
  04526626
12 Dec, 2015 11 commits
- Fixed old tests. Also added more max pooling tests. · b7e127f2
  Davis King authored Dec 12, 2015
  
  b7e127f2
- Added test for max pool layer. · 18d0f0f4
  Davis King authored Dec 12, 2015
  
  18d0f0f4
- Added max_pool_ layer. · df7d7f03
  Davis King authored Dec 12, 2015
  
  df7d7f03
- Fixed some resource leaks. Also fixed max_pool so it does exactly what the · 7ae43ae2
  Davis King authored Dec 12, 2015
```
spec says it should.
```
  7ae43ae2
- Made test_layers() a little more robust. · cbd57be6
  Davis King authored Dec 12, 2015
  
  cbd57be6
- removed cruft · 9065f08c
  Davis King authored Dec 12, 2015
  
  9065f08c
- Improved softmax tests · a4c5983a
  Davis King authored Dec 12, 2015
  
  a4c5983a
- Fixed bug in softmax gradient computation. · 0a832e42
  Davis King authored Dec 12, 2015
  
  0a832e42
- updated tests · 4aa0e3be
  Davis King authored Dec 12, 2015
  
  4aa0e3be
- Figured out the *undocumented* requirements for calling cuDNN's · ee6e54b4
  Davis King authored Dec 12, 2015
```
cudnnAddTensor() function and updated the specs and asserts accordingly.
```
  ee6e54b4
- Fixed race condition that could happen if set_size() was called while a cuda · 0babe27a
  Davis King authored Dec 12, 2015
```
kernel was still running.
```
  0babe27a
09 Dec, 2015 7 commits
- saving tests for add() · 0f840db3
  Davis King authored Dec 08, 2015
  
  0f840db3
- Improved cudnn error messages. · 3559f785
  Davis King authored Dec 08, 2015
  
  3559f785
- Made test_layer() more robust. · d1796934
  Davis King authored Dec 08, 2015
  
  d1796934
- Updated multiply()'s CUDA implementation to reflect it's new features. Also added · 2ba29f65
  Davis King authored Dec 08, 2015
```
CUDA version of add_bias_gradient().
```
  2ba29f65
- Added affine_ layer. · 2f34414e
  Davis King authored Dec 08, 2015
  
  2f34414e
- Added cpu version of add() and also added new add_bias_gradient() function. · 8062663c
  Davis King authored Dec 08, 2015
  
  8062663c
- Added missing assert check · 20d46fc5
  Davis King authored Dec 08, 2015
  
  20d46fc5
08 Dec, 2015 10 commits
- Made multiply() more flexible and also fixed a bug in the CPU implementation of · a29086bf
  Davis King authored Dec 08, 2015
```
batch_normalize_conv.
```
  a29086bf
- Made the affine_transform functions consistent. · 29f56b12
  Davis King authored Dec 08, 2015
  
  29f56b12
- Added the dropout layer · 76786430
  Davis King authored Dec 08, 2015
  
  76786430
- Clarified spec · 3222a3af
  Davis King authored Dec 08, 2015
  
  3222a3af
- Added another version of multiply() · 5f8e41a8
  Davis King authored Dec 08, 2015
  
  5f8e41a8
- Added bn layer tests. · 79344339
  Davis King authored Dec 07, 2015
  
  79344339
- Implemented the bn layer. · 7550681b
  Davis King authored Dec 07, 2015
  
  7550681b
- Increased default mini-batch size to 32. · 363b6b2f
  Davis King authored Dec 07, 2015
  
  363b6b2f
- Made the batch normalization code assign the parameter gradient outputs instead · b818b553
  Davis King authored Dec 07, 2015
```
of add to them so that it's consistent with how the layer interface expects
this to be done.
```
  b818b553
- Relaxed test_layer() a little. · 2c1e67f1
  Davis King authored Dec 07, 2015
  
  2c1e67f1
06 Dec, 2015 9 commits
- Added conv_ spec and did a little cleanup. · 81847660
  Davis King authored Dec 06, 2015
  
  81847660
- code cleanup · adc022c7
  Davis King authored Dec 06, 2015
  
  adc022c7
- Set initial bias parameters to 0. · 437da23e
  Davis King authored Dec 06, 2015
  
  437da23e
- Added a convolution layer. · 57a1072f
  Davis King authored Dec 06, 2015
  
  57a1072f
- Yet more robustness tweaks for test_layer(). · aa537dcc
  Davis King authored Dec 06, 2015
  
  aa537dcc
- More robustness tweaks for test_layer() · d08cd587
  Davis King authored Dec 06, 2015
  
  d08cd587
- Cleaned up the tensor_conv interface a little. Also fixed an error in the spec · 038c73d8
  Davis King authored Dec 06, 2015
```
for this object.
```
  038c73d8
- Made alias_tensor serializable. · 83ecf1d9
  Davis King authored Dec 06, 2015
  
  83ecf1d9
- Made test_layer() a little more robust. · ad40ddd3
  Davis King authored Dec 06, 2015
  
  ad40ddd3
05 Dec, 2015 1 commit
- Added GPU versions of the batch normalization functions. · cbce85ec
  Davis King authored Dec 05, 2015
  
  cbce85ec