search

vanderschaarlab / autoprognosis

A system for automating the design of predictive modeling pipelines tailored for clinical prognosis.
https://www.autoprognosis.vanderschaar-lab.com/
Apache License 2.0
114 stars 26 forks source link

Longitudinal Data Handler / Questions #25

Closed simonprovost closed 1 year ago

simonprovost commented 1 year ago

Dear Authors,

Wonderful system, thanks for this contribution! I have just came through the three articles and a video of the lab's director demonstrating this innovative AutoML BO system (https://www.youtube.com/watch?v=d1uEATa0qI).

However, something piqued my interest. Professor Van Der Schaar discussed Survival Models, Competing Risks, Temporal Models, and Causal Models at exactly 38:21 minutes of the video linked above. Where should I seek for Temporal / Longitudinal data handler in the code please? I am feeling like missing something to be honest after having read the paper that does not mention a lot about this..

I attempted to search through the GitHub repostiory: Temporal, Long, and Longitudinal, but nothing came up, I am guessing it'd be easier to ask for guidance or clarification,

Thank you very much for your help in advance, Best wishes,

DrShushen commented 1 year ago

Hi @simonprovost. This was likely in reference to the lab's other project - clairvoyance2 which is focused on time series data. Please note that that project is still a work in progress, and we do not yet have an AutoML component there, for instance. The "Survival Models, Competing Risks" aspects will be added in the next release, but do have a look at the models already implemented.

simonprovost commented 1 year ago

Okay, so to confirm, neither a Longitudinal handler nor a Survival / Competition Risks feature are available in AutoPrognosis. Nonetheless, Clairvoyance2 intends to do so in the coming releases, or at least for the Survival Models and Competing Risks you mentioned; longitudinal data may not be of use in this setting, right for all this?

If so, that is excellent; otherwise, additional clarrification would be welcome.

Thank you so much for your promptness, Cheers,