-
Hardware at the core: Nvidia GPU
- Kepler eg K80
- Maxwell eg M60
- Pascal (wiki) eg P100
- Volta (to come)
-
System
-
- CUDA 7
- CUDA 8
-
Deep Learning frameworks (name, backed-by)
- tensorflow, Google
- keras, Google
- caffe/caffe2, Facebook
- torch/pytorch, Facebook
- mxnet, Amazon for AWS
- cntk, Microsoft
- Theano,
- dl4j,
-
Model architecture (link to keras application if there is one)
-
Special purposed architecture
- U-net (arxiv)
- V-net
- E-net
-
Building blocks
- densely connected layer (my tutorial: ann101)
- locally connected layer
- convolutional layer (my tutorial: cnn from scratch, my tutorial: minimal cnn for MNIST)
- pooling layer
- dropout layer
-
Techniques
- Optimization (a lot)
- Batch normalization
- Data augmentation (my tutorial, my tutorial using MNIST)
- Feature extraction using pretained weight (my tutorial: VGG and Xception)
- Transfer learning (my tutorial: predict dogs-vs-cats using ImageNet-trained Xception net)
- Visualization computational graph and training process (my tutorial: visualize computational graph, my tutorial: training and validation loss monitoring)
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|