Hi, by my previous post you might have guessed that I'm interested in coding some kind of CR system with NN's. Untill now i've only used them for bots in games and i'm not sure what would be the best way to start with CR. I had 2 kinds of ideas in mind: 1) A program that recognizes handwriting 2) A program that can recognize pictures of fonts - this would be used to turn scanned pictures into text files. Now option 2 would be more usefull (for me atleast) and it'd be easier to train the NN but i'd have to do the initial step of changing a picture of text into individual pictures of characters - how difficult is that? The second option wouldn't be as usefull but it'd be cooler - it also wouldn't involve any of the initializing steps used in 1. What's the best way to start? (it doesn't have to be one of the choices above) Oh, and i'll probably be coding in C#.net if that matters. Thanks.

Grab a book on OCR (Optical Character Recognition). Most modern tomes in this field cover the issues with using ANNs on this task (and the how-to as well). You'll get more info from one targeted book than you will from 10 pages of posts here! ;)

Most character recognition is done by NNs (they're really good at high noise data). If you'd like to try out some other Machine Learning algrithms, Weka is a free, open-source tool-kit for java. It says it's for "Data Mining", but its really more of a machine learning package.