Home › Neural Network Forums › Neural Networks › General Neural Networks ›

CTRNN

fifo_thekid's picture
Submitted by fifo_thekid on Sat, 07/17/2010 - 12:56
in

Dear Sir,

I've purchased both of your books about C# Neural Networks Programming
They simply the best. They were an invaluable source of information
I've used various types of ANNs found in your books to train robots on navigation using MRDS2008 R2
Unfortunately, I couldn't find any implementation for Continuous Time Recurrent Neural Networks
When I tried searching online, I didn't find any lectures or tutorials on how to model or program those networks
Can you point me in the right direction please?

Thank You

‹ Req. Help: Software Classification with SOM Breif RBF Implementation Question ›
jeffheaton's picture
Submitted by jeffheaton on Sun, 07/18/2010 - 19:10.

First of all, thanks! Very glad that you like the books.

This is the first I've heard of a CTRNN, I've done recurrent neural networks but have not done one that adds the continuous time element to it.

I did some searching, and I can't find any solid examples of it either. I did find this book that contained some information on them. This one defines it fairly well, but it is not at all a "light read".

http://books.google.com/books?id=Guz7Q7MT1SgC&pg=PA340&lpg=PA340&dq=Cont...

fifo_thekid's picture
Submitted by fifo_thekid on Mon, 07/19/2010 - 00:05.

Dear Sir,

Thank you for the fast reply

According to knowledge I could acquire from other resources, CTRNN is not only about continuous time domain
CTRNNs continue learning even without supplying new input
They work just like the brain which can enhance its knowledge by thinking only without doing new activities
Is there any other neural network that can do the same thing?

michiel van den anker's picture
Submitted by michiel van den... on Mon, 07/19/2010 - 05:53.

fifo_thekid said:
"They work just like the brain which can enhance its knowledge by thinking only without doing new activities
Is there any other neural network that can do the same thing?"

Sounds to me like a "Deep belief network"*, or try google'ing "deep learning neural net" on video.google.com for a nice Google Tech Talk on the matter.

*http://www.scholarpedia.org/article/Deep_belief_networks

fifo_thekid's picture
Submitted by fifo_thekid on Fri, 07/23/2010 - 15:18.

Thank you very much for the vital information

To Mr. Heaton:
I've read in another article that you are working implementing Deep Believe Networks in Encog
Is there any progress in this regard?
Can you give me some tips on how to implement them using Encog?

Thank You

jeffheaton's picture
Submitted by jeffheaton on Wed, 10/13/2010 - 18:38.

I have not done anything yet in Encog that would further adding Deep Belief/Restrictive Boltzmann Machines. It is definatly something I would/will get to eventually, other priorities have prevented thus far.

It does not look too complex of an addition. Encog already has a Boltzmann machine, and as I understand it, that is what most deep belief is based on.

It uses a unique training method, that is not currently implemented in Encog. It also sounds like it can train with backpropagation, or other propagation techniques.

If you are interested in implementing it I can look further and probably give some additional insights on how to add this to Encog.

fifo_thekid's picture
Submitted by fifo_thekid on Sun, 10/17/2010 - 02:54.

Currently I'm working remotely with a Japanese research institute which has research teams each one allocated a different task.
I find your framework promising and pretty good that's why I'll try directing their research into Encog by the time I'll be there, hopefully...

jeffheaton's picture
Submitted by jeffheaton on Mon, 10/18/2010 - 16:43.

Good luck, and if I can help the process, let me know.

Copyright 2005 - 2012 by Heaton Research, Inc.. Heaton Research™ and Encog™ are trademarks of Heaton Research. Click here for copyright, license and trademark information.