jeudi 9 avril 2015

Iterative training in NN toolbox (Matlab)


Vote count:

0




I am going to make a brief explanation so you can understand better my problem.


I have multiple sequences of 119 3D points. Each sequence of 119 3Dpoints corresponds to an epileptic seizure of a patient. In these sequences I have the first 72 points labeled as a zero (corresponding to a non-seizure zone) and the 24 last points labeled as a 1 (corresponding to a seizure zone). The rest of the points are not labeled.


I would like to use this labeled data to train a time-delayed neural network (NN toolbox matlab) and to test with some other sequences.


I have tried training my model iteratively with my different chunks of data but the results I get when I test my network are really bad and recognise everything as non-seizure zone(0).


Do you have any ideas on how I should proceed?



asked 25 secs ago







Iterative training in NN toolbox (Matlab)

Aucun commentaire:

Enregistrer un commentaire