Reference-Implementation of a Convolutional Neural Network from scratch using C++ without parallelization.
Used libraries: Armadillo, Google Test
1. Clone repository git clone https://github.com/felixp98/cnn_cpu.git
2. Install Armadillo version 8.4 or higher (see below).
3. Create a build directory, e.g. mkdir cmake_build
4. Update the correct path for the MNIST data in cnn_reference_test.cpp
5. cd into the build directory and generate build files: cmake ../
6. Build the project using make
7. Train and test the model using ./cnn_reference_test
- Install Armadillo prerequisities:
sudo apt install cmake libopenblas-dev liblapack-dev
- Download newest Armadillo version: http://arma.sourceforge.net/download.html
- cd into extracted directory
cd arma*
- Generate build files:
cmake .
make
sudo make install