python - DBN deep network does not recognize many classes -

i using python code dbn downloaded deeplearning.net ( http://deeplearning.net/tutorial/code/ ). using dataset pamap2 downloaded uci archives. follow following steps:

1. normalizing data set have no negative values
2. i pass dataset unmodified neural network
3. calculating confusion matrix on results

the confusion matrix on validation set is:

[[1816    0    0    0    0]  [   0  326    0    0    0]  [   0  298    0    0    0]  [   0  305    0    0    0]  [   0  426    0    0    0]] 

ranking first , second class, other not ranked well, how should do? there paper can read? other datasets (eg. mnist) network fine