Author 
Topic: How similar are two given functions to eachother? 

stanleyaskit 
Posted: 30Apr09 21:00 



Hi,
I need to develop a Neural Network Code (My supervisor has asked for it), and I think I am in a big trouble now. It's a matter of life and death!
I would be thankful if somebody can give me a clue of how to start, or where to start from: Book, Article; Any form of information is really appreciated!
I define the problem as follows:
"Suppse we have a series of given functions, lets say f_1(x) to f_n(x) by which we want to train a neural network. They are very close, like they have a peak at some point and the position of the peak on xaxis varies a little bit toward right or left in each one. In reality all of these functions are experimental results of some theoretical function f(x)  that's why they are similar to each other, they have to be.
After the neural network is trained, we will give a data set g(x) as input function to it. Then it should be able to tell you how close or similar the given function, g(x), is to those functions that we have trained the system with, like it says function g(x) is %95 similar to function f(x), or %5 different."
I'm a mechanical engineer... I have enough knowledge of programming with Matlab, C++. I've done numerical simulations before. So please let me know if there is anything difficult, I can handle it.
Thank you very much in advance... Stan 


punnoose 
Posted: 17May09 08:45 



if u know (x,y) mappings of some of the points of two functions, u can create two radial basis function network using that known points. RBFs are used as function approximators and is slightly advantageous over Back propogation nwtworks
Please refer Introduction to neural networks by christopher bishop 

