harvard-edge / cs249r_book

Collaborative book Machine Learning Systems

https://harvard-edge.github.io/cs249r_book/

Other

513 stars 102 forks source link

CTGan Synthetic Data generation #297

Open emmanuel2406 opened 1 week ago

emmanuel2406 commented 1 week ago

Ex. 5.3

This is for the external Colab notebook attached for that exercise.

Suggestion: Perhaps copying this notebook and changing a few things would make it a smoother experience for the user.

I have found two things:

Downloading the dataset patients.csv proves to be quite nontrivial - I bypassed the terminal command and simply just copied the file into my Google drive.
CTGanSynthesizer does not seem to exist in the latest version of ctgan. I have seen online references to the library sdv, which has this method, but I then have to change all the commands going forward.

profvjreddi commented 1 week ago

@emmanuel2406 Thanks for reporting this issue. Would you be able to share your Colab so that we can take a look at it? @shanzehbatool if you have some time to take a look at this it would be great!

emmanuel2406 commented 1 week ago

@emmanuel2406 Thanks for reporting this issue. Would you be able to share your Colab so that we can take a look at it? @shanzehbatool if you have some time to take a look at this it would be great!

Sure, seems like it wasn't serious though. To run the cells you must just replace CTGANSyntehsizer with CTGAN. Here is a copy of the notebook https://colab.research.google.com/drive/1jUBiyhW4_PaWt0RSGk2ewnBUDrnF2d6J#scrollTo=cpV1FWHevaWO

shanzehbatool commented 4 days ago

Just saw this; I'm not able to access the Colab shared here but yes, CTGAN should work instead of CTGANSyntehsizer. And would need to directly obtain the dataset from Synthea. If the Colab shared works, then great, else I could also look into an alternative.