mp2893/gram

Label

Opened this issue · 3 comments

pipihou commented

Hi,I have discovered that your training label and training feature are the same.Your purpose is to predict the code of the next visit , so I suppose the training label should be one time later than the training feature.

mp2893 commented

Hi Pipihou,

You are correct that the label should be one step after the input.
Note that I deal with that issue in "padMatrix" function.
Please see the code, and let me know if you think there is still a problem.

Thanks,
Ed

pipihou commented

Thanks. By the way, why do you use ".3digitICD9.seqs" file as the label ,the medical concept is encoded differently from that in the seq file generated from build_trees.py.

mp2893 commented

Incorporating domain knowledge (the diagnosis hierarchy) is for the embedding matrix only. You can actually predict whatever you want. I just chose to predict 3-digit ICD9 codes as an example.