Tensorflow implementation of paper: Learning to Diagnose with LSTM Recurrent Neural Networks.

excuse me，the original is a multi-label classification problem, I did not get to know how your label is processed, after study your code I think your label is a single multi-label classification problem, am i wrong? would you help me to explain the label question? very very thank you!!

Hello, Regarding the illustration of the shape of the data: "Each training example will be a sequence of shape [1, time_steps, number_of_features]" However, in the code: "segments=segments.reshape([len(segments),(win_size + 1),1])" If I understand correctly, win_size = number_of_features + 1 (because an extra feature of the tims step) and len(segments)=time_steps So it seems there is a mismatch between the order of dimensions in the illustration and the code

Could you please clearify this issue? Thank you!

First of all, thank you for the code!

I know it has already been discussed, but it is not clear to me the data format in the csv file needed to run your code. Does it have to be two coloumns, first coloumn with the full time series and in the second coloum the label associated with each time step? or what else?

Thank you!

hi,

Thank you for sharing the code. I got the following error when constructing the multilayer LSTM:

InvalidArgumentError: Dimensions must be equal, but are 128 and 65 for 'rnn/while/rnn/multi_rnn_cell/cell_0/lstm_cell/MatMul_1' (op: 'MatMul') with input shapes: [?,128], [65,256].

I am using tensorflow 1.12 on Windows 10.

I believe the issue is on how the multiple LSTM cells are in created. Specifically, this line multi_layer_cell = tf.nn.rnn_cell.MultiRNNCell([cell] * 2) likely creates two cell pointing to the same object. I was able to get rid of this issue by changing it to: multi_layer_cell = tf.nn.rnn_cell.MultiRNNCell[rnn_cell.LSTMCell(n_hidden, state_is_tuple=True) for _ in range(2)])