simoninireland, 9 months ago

@gvwilson My colleague Tom Kelsey, not on Mastodon, is an expert in this with respect to evaluating machine learing applied to medical data https://www.st-andrews.ac.uk/computer-science/people/twk

gvwilson, 9 months ago

@simoninireland thank you

almenal99, 9 months ago

@gvwilson you can check the Synthetic Data Vault for Python

https://sdv.dev/

If you click on the squares at the top of the main page (the ones that say 'Single Table', 'Multi Table', etc) it will open a Google Colab notebook with tutorials

SherlockpHolmes, 9 months ago

@gvwilson This is a review paper: https://krisrs1128.github.io/LSLab/assets/papers/generative_review.pdf and the code is all here: https://github.com/krisrs1128/generative_review

going_to_maine, 9 months ago

@gvwilson the scikit learn docs are decent for this, if I recall correctly.

gvwilson, 9 months ago

@going_to_maine thanks

benschneider, 9 months ago

@gvwilson The “synthpop” R package is excellent for when you have an existing dataset and you want to create a synthetic version (e.g., for sharing data while protecting research participants’ confidentiality)

gvwilson, 9 months ago

@benschneider thank you

rwxrwxrwx, 9 months ago

@gvwilson Markov chain Monte Carlo or bootstrapping might help you, depending on what you have in mind.

gvwilson, 9 months ago

@rwxrwxrwx thank you - I'm looking for tutorials or packages with tutorials right now

rwxrwxrwx, 9 months ago

@gvwilson FWIW, TensorFlow Probability and PyMC both implement Markov chain Monte Carlo (although it's not hard to write your own) and have documentation available.

omearabrian, 9 months ago

@gvwilson https://github.com/ropensci/charlatan could be good #rstats resource. @rOpenSci

gvwilson, 9 months ago

@omearabrian thank you