mp2893 / mime Goto Github PK

https://github.com/sjy1203/GAMENet

This is not an issue but i have a question about extracting the dx embeddings from the model.
I created a model till the dx embedding layers.
Now, while creating the pickle files for the input, I have a dictionary which maps each diagnosis code with its respective integer(starting from 1 till the max number).

Now, in the embedding layer, can i directly extract the weights and map the index to the diagnosis code index? I noticed that the input is not shuffled with respect to the diagnosis codes but at patient level only.

Please tell me if there is a better way to do that.

I read your paper, its a great effort in Multi-level Medical Embedding. Can I know when detailed instructions will be added in readme file? Currently it is quite difficult to understand just the code. Instruction similar to ones that is in your Retain and med2vec and DoctorAI that shows how to prepare input data and train model will help users to use this work. Thanks!

please upload sample data and input data structure. It is very difficult to understand from code Without comments

Hi, I found your paper quite interesting and have a great scope in medical prediction.
Can you please share code in TensorFlow 2.0 for ease in understanding.

Hi! When will the source code of GCT be released？My diagnosis code doesn't lead to ordering specific medications.

in project.py there's this library
you didn't add that
("import structured_learning_util as slu")

Hi I really enjoyed your paper and I was wondering if you had an example of the files you used or how to apply this code on MIMIC.

Thanks.