Hello Uzma
I have used ANN for my postgraduate research way back in 1990. For simple understanding it is generation of non-linear regression mathematical equation where you provide multiple inputs and get single or multiple outputs. Once you establish the equation, all you need is to use that in your program. My research was limited to single output.
In order to obtain the equation, you would require actual data with inputs and output. Data with more variance is helpful so it can cover all possible cases. Note that ANN is really helpful where it is difficult to establish impact of Inputs on Output. Back-Propagation method is easy to train for such application.
As suggested by other readers, you can google , if you want to read my people , here is the link. http://people.inf.ethz.ch/fcellier/Pubs/FIR/aimed_96.pdf
Good luck with the project.
Habib Rehman
Buraq Group
From: pakgrid@yahoogroups.com [mailto:pakgrid@yahoogroups.com] On Behalf Of Uzma Naeem
Sent: Sunday, 26 August 2012 12:47 PM
To: pakgrid@yahoogroups.com
Subject: [pakgrid] Question for Artificial Neural Network (ANN) Experts
Dear ANN experts,
In one of my projects I need to use Artificial Neural Networks (ANN). I need to study the subject in depth but could you please give some quick info. regarding the following:
How ANN works? I mean how do we train ANN? For example if we train an ANN on some pattern (e.g. image) for a given case do we need to train again the network when we deal a new similar case or could be save this training information and use it when needed later? In short could we remember training information or we need to retrain every time we need to compare a given pattern with another pattern?
Could you please comment urgently to clarify the above?
Best,
Uzma
No comments:
Post a Comment