ghuw/simple-neural-net

Simple Implementation of 3 Layer Fully-connected Neural Network in Python

Simple Neural Network for MNIST Dataset

• Simple 3 Layer Neural Network in Python by using only basic library (Numpy).
• Sample multiclass using MNIST data provided by http://yann.lecun.com/exdb/mnist/

Getting Started

git clone https://github.com/ghuw/simple-neural-net

Prerequisites

• Python3
• 1. Linux

sudo apt-get update && sudo apt-get install python3.6

• 2. Windows https://www.python.org/downloads/

• Numpy

pip3 install numpy

• Dill

pip3 install dill

• Matplotlib

pip3 install matplotlib

Usage

Download and extract dataset from yann.lecun.com/exdb/mnist/ and move to working directory.

• train-images-idx3-ubyte.gz: training set images (9912422 bytes)
• train-labels-idx1-ubyte.gz: training set labels (28881 bytes)
• t10k-images-idx3-ubyte.gz: test set images (1648877 bytes)
• t10k-labels-idx1-ubyte.gz: test set labels (4542 bytes)

Running the classifier

python3 main.py

Author

• ghuw

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Acknowledgments

• This project is inspired from miloharper work.