This code is a simple application of the most basic neural network.
The goal is to distinguish images with persons in them.
The images are from home security cam mounted at a fixed position. 
Changes in shadow usually triggers the security cam.
This simple code is to build a neural network model to find photos truely has persons in them, not just shadows or sometimes a bird flying by.

The code is modified from Stanford Machine learning online class (orignally in Octave).
Since I have not learned how to use write/use an advanced gradient decent algorithm yet the code contains a barebone gradient decent code.