Deep Learning framework using C++ CUDA Currently only contains the C++ impementation. This is how it works:
The following layers are defined: Linear, ReLU, Sigmoid The following losses are defined: MSE
Create pointers of type layer with the right parameters, then create a model object, adding the layers to the model using model.add()
All operations are done on a data structure called Matrix. Source code provided To run the test code, use these commands called from the directory where dump.cpp is stored (all the files need to be in the same directory). It runs a 2 layer network on 1000x2 datapoints for binary classification
g++ -o output.exe dump.cpp
./output.exe