Today I have started looking through neural networks and CvANN_MLP. For the training, I generate one big matrix with information for the training samples. Is possibly to: Generate matrix for one sample at a time, train the neural network, save it in XML, load it and continue the training ?
no, not possible.
you can neither cut your training data into cute little blocks and apply them one by one, nor can you append new items, once the thing is trained. it tries to find an optimal model for your data, so it needs to see all of it at the same time.
same applies to probably all ml-algos in opencv.
Thanks. ;) And an other question: is there a way to get the progress of training procedure ?
yes.In the CvANN_MLP:: train method you have to set flag parameters with UPDATE_WEIGHTS.
oh.
@Mostafa Sataki, are you going to make an answer from that(so ppl can upvote) ? or do i have to edit, and prove myself wrong ?
Asked: 2013-09-09 07:07:59 -0600
Seen: 555 times
Last updated: Sep 09 '13
