tyler-barham / number-predictor

An implementation of a neural network predicting handwritten numbers.

Number predictor

This program demonstrates a neural network using backpropagation to predict handwritten numbers. The input images are 20x20 pixels, and the output will be the prediction of the image ranging from 0-9. The neural network is fully connected with one hidden layer and has a shape of 400x25x10 (excluding bias neurons).

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

GNU Octave, version 3.8.0

Installing

$ git clone https://github.com/Tyler-Barham/number-predictor.git

Usage

$ cd ./number-predictor
$ octave-cli
$ main

Example output

Figure 1. A sample of 100 images from the dataset.

Figure 2. The neural network predicting the number in an image.