When am trying to train data using nn toolkit in matlab am getting error as 0 and when i checked in nntrain tool validation check is at 0 and am getting notification as minimum gradient reached. Can someone help is what results am getting is correct or not? since am getting 100% accurate result. but validation check is 0. Please someone answer it. Thanks in advance.
I am not getting how you are sure about 0 error while training. I think your NN is not trained perfectly. You may increase the number of epochs and give more training to the ANN and do the simulation again. If the training is 100% done, then simulation results must give good accuracy.
I think you are facing the problem of Over fitting (also called High Variance). It is the phenomena in which a classifier works very well on train data, but very bad on test data.
Following link can be very useful
https://www.coursera.org/learn/machine-learning/lecture/ACpTQ/the-problem-of-overfitting
hi ANIL
something is not done well in training youre ANN. first dont use same input data for training and testing. second use more epochs and use more data for training.
Good luck
Did you really shuffle well? Your training data may consists of training data of only one class and you test data of another class which your neural net never has been trained for.
Check also your labels. Are they correct labeled?
- Badges
- Science topic