This repository implements some neural networks from scratch using python programming language. Currently master branch contains multi layer perceptron (MLP) neural network which trains by backprobagation algorithm using gradient descent optimizer and convolutional neural networks based on keras module. In the GPU branch, we try to use benefits of nvidia's cuda toolkit to accelerate the learning process.

We recommend to use a virtual environment to install this package. This is a simple guide to install this package.

git clone https://github.com/alirezaafzalaghaei/neural-netwroks-from-scratch.git
cd module
pip instal pipenv
pipenv install
pipenv install -e .
cd examples
pipenv run python iris_test.py 

see main.pdf file in report folder for detailed information about this module.

• Porting MLP for Nvidia GPUs.