This set of MATLAB functions implements a two-layer neural net to identify hand-written text to up to 99% accuracy.

The main.m function runs through the entire training and and testing process for a preloaded set of data (contained in data.mat). To modify the data input, you can set this file in the main.m section labeled "Loading and Visualizing Data". The minimization function used is Carl Edward Rasmussen's 2001-2002 fmincg function, which is well-suited for fast and accurate minimization of low-dimensional data sets. Classification is done using a multiple-output sigmoid classifier (implemented in sigmoidGradient.m).

I credit a large portion of the training data and some of the image visualization code in main.m and displayData.m to Andrew Ng's online Coursera course's homework #4 (found at https://www.coursera.org/learn/machine-learning).