does this model work ?

[bicepjai]

I didnot use mxnet. i did implementation on keras and this is the only model that gives the worst performance even below the benchmark models like naive bayes. i thought there must be some bug in my code, but after seeing your comment. i would question the model too. The model details is provided below. id you can understand it. please comment on it

[Ldpe2G]

@bicepjai The output shape of the each layer is strange, could you provide the code of the network definition in Keras? So that I can tell whether you have implemented the network right.
The performance of AC-BLSTM is confirmed by many exprtiments. What I meant in readme is that If you just run this code on one of the datasest meantioned in the paper without any modification, it may not reproduce the results. Because when I conduct the experiments I use some tricks during traning, for example early stopping and change the learning manully, try different dropout rate, or activation functions and so on.