AForge.NET

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

Problem learning ANN

Forum to discuss AForge.NET Framework, its features, API, how-tos, etc.

Problem learning ANN

Postby nattsurf » Fri Jun 05, 2009 12:50 pm

What could I try if the alpha value doesn't help in training the neural network?
I figured out that if I add more rows to the input vector array then it is harder to reach below 1.
Also another thing. If I have a vectors that have many -0.5 values at the end like this:
r= random double number range -1 to 1
r,r,r,r,-5,-5,-5,-5,-5,-5,-5
r,r,r,r,-5,-5,-5,-5,-5,-5,-5
r,r,r,r,-5,-5,-5,-5,-5,-5,-5
r,r,r,r,-5,-5,-5,-5,-5,-5,-5
Then it becomes harder to reach below 1.

But this improves if I change to.
r,r,r,r,-5,-5,-5,-5,r,r,r
r,r,r,r,-5,-5,-5,-5,r,r,r
r,r,r,r,-5,-5,-5,-5,r,r,r
r,r,r,r,-5,-5,-5,-5,r,r,r

Sorry for my lack of understanding in this area :)
nattsurf
 
Posts: 6
Joined: Fri May 22, 2009 11:16 pm



Return to AForge.NET Framework