These 27 commits are when the Protocol Buffers files have changed:
| Commit: | b5da247 | |
|---|---|---|
| Author: | milakov | |
Migrated to VS2017
The documentation is generated from this commit.
| Commit: | 831ee44 | |
|---|---|---|
| Author: | milakov | |
Extended ReLU to Leaky ReLU: optional negative_slope added
| Commit: | 5a82c3d | |
|---|---|---|
| Author: | milakov | |
Dilation is supported in convolution layer
| Commit: | 3de85c2 | |
|---|---|---|
| Author: | milakov | |
Per feature map mode added to dropout
| Commit: | 44515bc | |
|---|---|---|
| Author: | milakov | |
Strides added to sparse convolution layer
| Commit: | 4b0b503 | |
|---|---|---|
| Author: | milakov | |
Fixes to ST layers
| Commit: | 0eb8e2d | |
|---|---|---|
| Author: | milakov | |
affine_grid_generator_layer added (for Spatial Transformer)
| Commit: | 7794626 | |
|---|---|---|
| Author: | milakov | |
Strides added to max subsampling layer
| Commit: | aab4b7b | |
|---|---|---|
| Author: | milakov | |
Added round_up mode for max subsampling layer
| Commit: | 4882fd5 | |
|---|---|---|
| Author: | milakov | |
Average subsampling layer now allows specifying output size instead of subsampling size
| Commit: | 43c15bc | |
|---|---|---|
| Author: | milakov | |
1) w/out bias mode added to convolution layer 2) check_gradient command added 3) batch_norm layer added
| Commit: | c2021ac | |
|---|---|---|
| Author: | milakov | |
1) Stride added to convolutional layer, 2) Alpha added to Add layer
| Commit: | b3f12f8 | |
|---|---|---|
| Author: | milakov | |
1) CDF2PDF layer added, 2) EntryConvolution layer added, 3) alpha can be specified for AverageSubsampling
| Commit: | bf6774e | |
|---|---|---|
| Author: | milakov | |
Upsampling layer added
| Commit: | 73546ed | |
|---|---|---|
| Author: | milakov | |
CDF Max layer added
| Commit: | 64b57fa | |
|---|---|---|
| Author: | milakov | |
Added "min" function to MaxSubsampling layer
| Commit: | 7a335ce | |
|---|---|---|
| Author: | milakov | |
Average and Max subsampling layers are now capable of subsampling in feature map and entry directions
| Commit: | cc443f6 | |
|---|---|---|
| Author: | milakov | |
MSE layer reworked into generic LError layer (L2 by default)
| Commit: | 9adefd9 | |
|---|---|---|
| Author: | milakov | |
Reshape layer added
| Commit: | 814f123 | |
|---|---|---|
| Author: | milakov | |
Error functions accept the 3rd optional input layer - mask
| Commit: | 0d456b4 | |
|---|---|---|
| Author: | milakov | |
Imagenet training is working now
| Commit: | 7ad2d8a | |
|---|---|---|
| Author: | milakov | |
Inference is working for GTSRB
| Commit: | 0ff5b2c | |
|---|---|---|
| Author: | Maxim Milakov | |
Sigmoid layer is allowed to affect selected feature maps
| Commit: | 646c5a4 | |
|---|---|---|
| Author: | milakov | |
Overfeat functionality added (see tiling option of max subsampling layer, and untile layer)
| Commit: | 3e20f79 | |
|---|---|---|
| Author: | milakov | |
Input and output data normalizers are stored in protobuf format now. Use convert_input_normalizer and convert_output_normalizer to convert existing binary normalizers to new format.
| Commit: | cb08d1b | |
|---|---|---|
| Author: | milakov | |
File schema moved to protobuf. Use convert_schema action to convert existing schemas.
| Commit: | 99949eb | |
|---|---|---|
| Author: | milakov | |
Compiles in Linux and Windows with Protobuf lib