Before and during graduate school, I was into adaptive programming, particularly neural networks (NN).
I tried to incorporate NN programming in several projects for classes.
2. My Projects
I wrote this very simply recognizer for CS651 (Distributed Interactive Simulations).
It's an Excel spreadsheet (and works in OpenOffice 2.2.x).
To get the random numbers to change, simply edit any empty square.
- Trav's Tic-Tac-Toe Trainer (TTT) uses traditional (top-down) AI to train neural networks how to play Tic-Tac-Toe. It employs genetic algorithms to find optimal NN architectures. It's implemented in C++ and consists of source code only (no executables).
3. Good Online Resources
Introductions and References:
- Statsoft's NN Introduction
- This is a very attractive site with lots of pictures and a huge glossary. Excellent both as an introduction and reference.
- Backpropagator's Review
- A backpropagation "reading list" - links to dozens, maybe even hundreds, of online articles and papers. An advanced reference.
- Leslie Smith's Introduction to NN
- This is not as newbie-readable as the Statsoft pages, but contains some nice pictures and motivations for using NN.
- AI FAQ: Neural Nets
- A huge source of info, both introductory and reference. Warning: no pictures and somewhat dry.
- Multi-Layer Perceptron (Little Red Riding Hood)
- Backpropagation network, small and easy to understand. Implemented in C, a little bit lacking in comments or coding clarity, but overall a good introductory example.
- What a Neural Network Can Learn about Othello
- This is math-heavy, with no actual source code. Try the "PDF" link in the upper right.
- NNUGA - Neural Network Using Genetic Algorithms
- General recognizer-type program implemented in C. This example is pretty dry, but the code is easy to understand. To actually run it, you also need Tk.