stop backprop-learning

• Messages sorted by: [ date ][ thread ][ subject ][ author ]
• Next message: : "Re: Problems with Rprop-Algorithm in batchman-mode"
• Previous message: : "ART2 result file"

Hello SNNS-Users!

How can I stop the training of an Elman-Network (learning-algor. 
JE_Backprop) by a specified variable???
I want to stop the training, if the RMS-Error of validation or the 
MS-Error of the trainig is less than 0.05 or 0.1. I have set the 
variable "dmax" in the conrol-panel to 0.05 or 0.1, but the training 
doesn´t stop by reaching this value! (sometimes it does it, but 
sometimes the error is 0.000 still end of training, that means during 
4000 cycles, but the error-graph reach this limit after 1000 cycles). 
Can someone exlpain me that??
Thanks for help

many greetings
Christoph Lange

______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com

• Next message: : "Re: Problems with Rprop-Algorithm in batchman-mode"
• Previous message: : "ART2 result file"