The library gave me the 1987 version of the book where they put color televisions on a pedestal (sp?). I think I will have to buy this book from ebay, there's just no other way around it. If I'm doing this I am going to have to shell out the big bucks.
Got some Matlab code off the Internet that does a Neural Network with two hidden layers. It does cropping and normalization on a bitmap and trains on it for enough epochs such that error is below some threshold. Then tests on another set of inputs. So far it seems it is working! Oh yeah, it's regarding a character recognition project for my AI course. I will have to change the input, because my training set is already normalized and is not in a bit map form. I hope this makes my life easier. Also, I only need one hidden layer. I wonder if I have to rewrite the feedforward code that Matlab provides in the NN toolbox. Oh well, one step at a time
Something works!!! Step one taken! Happy!


