This repository implements a simple MLP in three languages: Python, C++ and CUDA.

Python

It is recommended to use conda to install the required dependencies.

$ conda env create -f env.yaml

To create the train and test datasets, run make_csv.py

$ python make_csv.py

To run backpropagation in Python, run network.py

$ python network.py

C++

To build the C++ version, use make.

$ make run_cpu

You can also run tests using make.

$ make run_cpu_tests

CUDA

To build the CUDA version, use make.

$ make run_cuda

You can also run tests using make.

$ make run_cuda_tests