I have to say this, but a neural network without hidden layers is a bit... well, pointless, really . Make sure that it handles the "xor" problem too (if your example is the xor problem, should 1, 1 not give 0? Assuming that's what you're trying to do, of course)...

About the Delphi container classes: yes, this is a pain (I really wish that Borland would do generics some time soon!). In the meantime, try this library over at sourceforge, which might be interesting:

http://sourceforge.net/projects/decal/

Not perfect but it is a good start...

I might have to deal with good ol' overhyped XML soon . Anyone here have experience with XML and Delphi? If so, please pipe up.