:: AForge.NET Framework :: Articles :: Forums ::

help, train network

The forum is to discuss topics from different artificial intelligence areas, like neural networks, genetic algorithms, machine learning, etc.

help, train network

Postby jarain78 » Fri Nov 04, 2011 1:28 am

I`ve work out some things but now I`ve another issue:
I have to do a subroutine that in my case I have been called it , Crear_Neurona (), where I create my neuron and initialize hit this way:

Code: Select all
inputLayer = new LinearLayer(256);
hiddenLayer = new SigmoidLayer(10);
outputLayer = new SigmoidLayer(10);
new BackpropagationConnector(inputLayer, hiddenLayer);
new BackpropagationConnector(hiddenLayer, outputLayer);
Neuronas_Ocr = new BackpropagationNetwork(inputLayer, outputLayer);

All is in a sub routine, now I have another sub routine where I train my network neuronal.

My other problem is that when I do this the net training is to long and I don`t know if that`s normal, I asume that is becouse I`ve got ten output and 999 numbers to train the network and the time of training is relative or there is a way to make it longer or slower?
I calculate the main square error with 100 cicles but for training two numbers it`s taking to much time, hours and I`don’t know if that`s normal. Of course is to take in consideration that I`m talking about 100 images by number.

Thanks for the help

Posts: 6
Joined: Sun Oct 09, 2011 8:58 pm

Re: help, train network

Postby Tievoli » Thu Nov 08, 2012 3:48 pm

while neural networks are great, they do cost computing time.
Maybe you should spend more time to explain your question in detail.
After reading it i got no clue of what your trying to ask, what task you try to do, how you program looks.
how do you think people could answer you ?
Asking good questions results in good answers
Posts: 16
Joined: Fri Nov 02, 2012 1:48 pm

Return to Artificial Intelligence