Back Propagation Neural Network


I've always been interested in artificial intelligence techniques; one of the things we had discussed in my college AI course was neural networking. So I opted to create my own.

The project

Written in C++ using Visual Studio, the back propagation network is a simple implementation of a neural network. It allows you to specify layer size, and number of layers, and inputs/outputs. Other bits included a trainer program, as well as a set of tests to run to verify that the code was correct enough that the network could memorize a simple XOR data set. Methods exist for writing the network to disk, and reading it back, although these methods store the data in a text format, which isn't ideal for real use (read as: files are much larger than they need to be)

My role

This was a random solo project, and thus it was all me.

The results

The only thing I have used this for was a for a private add-on for a game, which could detect when certain events were happening (based on polling the state of the screen and running it through the network) and reacting to them accordingly.

The code

Code is available.