Problem of neural network training with CvANN_MLP

asked 2015-01-27 16:30:31 -0600


I am training some HOG features (729 features per example) with CvANN_MLP. It looks like the size of training set is limited for CvANN_MLP. If I pass 3000 examples to the train() function, it works fine. But if I feed 6000 or more examples, the train() function will return immediately with return code 0.

What is the problem?


edit retag flag offensive close merge delete


If I use RPROP, it works. But BACKPROP fails.

Lei Wang gravatar imageLei Wang ( 2015-01-29 13:52:32 -0600 )edit