i'm doing my final year project at university on machine learning algorithms for use on a simple game of tic-tac-toe and i am using Prolog with the XPCE add-on for the GUI.
I just wanted to know if it is possible to code a multi-layer perceptron with the backpropagation algorithm? it's a coursework assignment for my neural networks course which states we must use c++, java or perl but i hate them all!
im also studying neural networks but my lecturer absolutely sucks (he can't speak english and he just reads the slides out without turning once to face the students lol. His average is 41 slides per 30 mins!)
just hoping someone out there can help!!!
You can most definately use backpropagation on a multilayer perceptron. In fact, that is the standard way of training the network.
I am currently developing some code in Java for an article or set of articles on neural networks. If you have any more detailed questions, dont be afraid to ask.
in the algorithm on the backward pass when u are calucating the error values, new biases and new weights, i dont understand the symbols in the formula to update the new weights with the learning rate!
could u help me please?
Well there are many ways to do this, and many symbols to use to represent what is going on, so you should really post the symbols to be more specific.
However I can still make a guess at some of what you might be having trouble with.
I link to a couple of images that provide a common description.
They are both referenced from http://www-ra.informatik.uni-tuebingen.de
This is an inherent problem with mathmatitions devising these rules, they like to use one letter variables all the time. I will personally spell things out a little clearer. With good use of pseudocode.
Nomad, consider writing the articles for DevMaster