Neural Network fun

Setting up

0. For best experience, make sure you are on Ubuntu environment.

1. Make sure you have all the following dependencites installed:

   • python3
   • cmake version >= 2.8.5
   • libeigen3-dev

2. Clone this repo.

   git clone https://github.com/stanford-stagecast/nnfun.git

3. The rest should be straightforward:

   cd nnfun/
   mkdir build
   cd build/
   cmake ..
   make -j$16  # make -j$(nproc)
   make check # should pass the 3 tests
   

How to Run

0. Head to directory build.
1. Run cmake ...
2. Run make.
3. Run ./src/frontend/<file_to_run>.

File Stucture

• src: the directory containing basically all codes for the neural network
  • nn: the directory containing files of a generic neural network (please read the file, class, function descriptions in the files)
    • layer.hh: atomic element in this neural network
    • network.hh: network build using classLayer
    • neuralnetwork.hh: more higher level class directly providing gradient descent functionality
  • frontend: files using the neural network to predict
    • CMakeLists.txt: containing files to compile (feel free to comment out some files and add more files)
    • predict_inverse_256.txt: the file containing codes to predict 1/x in domain [1,100] (feel free to play with it)
  • test: testing files used in make check command (feel free to add more)
• docs: the directory containing some documentations and thoughts
  • how_to_use_neuralnetwork.hh.txt: the txt file containing some useful methods in neuralnetwork.hh
  • data_structure_ideas.txt. and tempo representation.txt: some previous ideas on note representations
• progress_reports: the directory containing some progress at certain time
  • 06_14_2022.txt: latest progress before summer 2022