Code does not work

[rfali]

While I highly appreciate the authors of the paper to have made the code available, and a very lengthy ReadMe file to go with it, unfortunately it does not work as intended. Some suggestions

1. The Heart Disease dataset should be included in the repo (it is after all public).
2. The Hypertenstion dataset is not included, and as such the code associated with it should have been removed (or the dataset included).
3. The file structure is not correct. The files have been moved into src directory but the figure1.py cannot access the necessary imports. Necessary changes like:
   import src.hierarchyrl.powerset, src.hierarchyrl.utils, src.hierarchyrl.policy2
4. Readme mentions that pkl files have been included or provided, but that is not the case.
5. I was able to produce Fig1 but it was not "reproduced" i.e. the image was different from the one in paper, while using the default settings.

I would highly suggest the authors/developers to be a little more prudent while making public their research. While the code is definitely welcome, it should be tested before made public.

[gbernardino]

Hi @rfali ,

Thanks for the interest. I apologise for the problems, but please be aware that it is a considerable amount of work to have an issue-free code that works in other's computer. This code was meant as support for reviewers, and I wanted to make it public for transparency and open science; it does not aim to be a finalized product. Again, I have to apologize since I wanted to improve some of the things, but got distracted with other projects. I am working on an extension of this, which I wanted to upload as well as a more definitive version, but it is lasting more than I anticipated.

1. The fact that it is public available somewhere doesn't always imply that I can simply put it in my own repository. I will check the license of the specific dataset .
2. This is more complicated, since it is clinical data. I am trying to determine with clinicians if derived indices are OK to be shared publicly (maybe after corrupting them with some noise). I think that even if I cannot upload the data
3. I just realised that the setup.py was missing. The code was supposed to work as an installed library. Now it should work with the correct imports (as long as you run "pip install -e .")
4. This is the same as #2, the .pkl data were supposed to be the HTA dataset.
5. I am very interested in this issue. Could you share the plot that you got?