[help] error when using fann

[Jiyu Zhang]

Hi, I am using fann2 recently and have met a problem.
when I am training the network, it has no extra screen-output for 3 hours,
and it doesn't stop executing.
The last two lines are:

Neruons 68. Current error: 0.371939. Total error:44694.3711. Epochs
15484. Bit fail 77607. candidate steepness 0.25. function
FANN_ELLIOT_SYMMETRIC
Neruons 69. Current error: 0.313447. Total error:37665.

It seems to me that it has met something wrong when it output the Total
error.

Has anyone ever met the similiar problem?
What reasons might it be?
Thanks a lot :)

[Adrian Spilca]

Hi Jiyu,

I would guess that your network fails to learn to an error you
requested. Could be that the problem you are trying has no solution
(your training set is not consistent) or more likely your training
process is stuck in a local minimum. To avoid this you might want to
change the architecture (increase the number of neurons, change output
function...), change the learning algorithm, try another initialization
of weights... Try google for "avoid local minima" for more advice on this.

http://www.faqs.org/faqs/ai-faq/neural-nets/

Regards,
Adrian

Post by Jiyu Zhang
Hi, I am using fann2 recently and have met a problem.
when I am training the network, it has no extra screen-output for 3
hours, and it doesn't stop executing.
Neruons 68. Current error: 0.371939. Total error:44694.3711. Epochs
15484. Bit fail 77607. candidate steepness 0.25. function
FANN_ELLIOT_SYMMETRIC
Neruons 69. Current error: 0.313447. Total error:37665.
It seems to me that it has met something wrong when it output the
Total error.
Has anyone ever met the similiar problem?
What reasons might it be?
Thanks a lot :)
--
Best Wishes
Jiyu
------------------------------------------------------------------------
-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
------------------------------------------------------------------------
_______________________________________________
Fann-general mailing list
https://lists.sourceforge.net/lists/listinfo/fann-general

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

[Steffen Nissen]

If there does not exist a solution, the cascade2 algorithm would still be
adding neurons, I suspect that you have encountered some kind of bug
(infinite loop or something like that), why only half the line is printed I
really do not know. If you can reproduce this error, please try to reproduce
it in a debugger and send a stack trace.

Best Regards,
Steffen

Post by Adrian Spilca
Hi Jiyu,
I would guess that your network fails to learn to an error you
requested. Could be that the problem you are trying has no solution
(your training set is not consistent) or more likely your training
process is stuck in a local minimum. To avoid this you might want to
change the architecture (increase the number of neurons, change output
function...), change the learning algorithm, try another initialization
of weights... Try google for "avoid local minima" for more advice on this.
http://www.faqs.org/faqs/ai-faq/neural-nets/
Regards,
Adrian

Post by Jiyu Zhang
Hi, I am using fann2 recently and have met a problem.
when I am training the network, it has no extra screen-output for 3
hours, and it doesn't stop executing.
Neruons 68. Current error: 0.371939. Total error:44694.3711. Epochs
15484. Bit fail 77607. candidate steepness 0.25. function
FANN_ELLIOT_SYMMETRIC
Neruons 69. Current error: 0.313447. Total error:37665.
It seems to me that it has met something wrong when it output the
Total error.
Has anyone ever met the similiar problem?
What reasons might it be?
Thanks a lot :)
--
Best Wishes
Jiyu
------------------------------------------------------------------------
-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
------------------------------------------------------------------------
_______________________________________________
Fann-general mailing list
https://lists.sourceforge.net/lists/listinfo/fann-general

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Fann-general mailing list
https://lists.sourceforge.net/lists/listinfo/fann-general