Adapt a neural network

[mehdiesteghamat]

Hi All

Hope you are all fine.

I am new to the OpenCV. I am trying to adapat a trained NN. I used
method "int CvANN_MLP::train" but it actually resets the weights as it
seemed to me.

How can I adapt a trained NN with a new set of training data?!

Best Regards,
Mehdi



Change settings: http://www.yahoogroups.com/mygroups, select
Get Emails (get all posts)
Daily Digest (one summary email per day)
Read on the web (read posts on the web only)Or Unsubscribe by mailing OpenCV-***@yahoogroups.com

[mehdiesteghamat]

Hi All

I found the solution. I should call the "trian" method with flag
'CvANN_MLP::UPDATE_WEIGHTS' like below:

train( &trainDataXFeature, &trainDataXResponse,
0, 0, nnEstimatorParamsX,
CvANN_MLP::NO_OUTPUT_SCALE | CvANN_MLP::NO_INPUT_SCALE |
CvANN_MLP::UPDATE_WEIGHTS);

but new ploblem raised. I get the flowing weird error:

"Some of new output training vector components run exceed the original
range too much"

In my contention, I should not get this error because I called train
class with no regards to input and output scales (using flags:
CvANN_MLP::NO_OUTPUT_SCALE | CvANN_MLP::NO_INPUT_SCALE)

I checked the OpenCV codes too. It seemed there exists some bugs in
the code??!

Could anybody help??

Best Regards,
Mehdi

Post by mehdiesteghamat
Hi All
Hope you are all fine.
I am new to the OpenCV. I am trying to adapat a trained NN. I used
method "int CvANN_MLP::train" but it actually resets the weights as it
seemed to me.
How can I adapt a trained NN with a new set of training data?!
Best Regards,
Mehdi

Change settings: http://www.yahoogroups.com/mygroups, select
Get Emails (get all posts)
Daily Digest (one summary email per day)
Read on the web (read posts on the web only)Or Unsubscribe by mailing OpenCV-***@yahoogroups.com