this is a small and pointless protest but I’ve had a COVID paper under review for the last 5 months with zero movement, so I’m just not reviewing other COVID papers any more

Show thread

getting constantly gaslamped by journals sending me “reminders” to review papers I never agreed to review (I.... think???)

food 

vanilla soy milk is so GOOD what the HECK

returning to my colleagues in machine learning to regretfully inform them that the human body is a series of tubes

Show thread

rainy bank holiday monday so I'm binge-watching lessons on ICU nursing + upgrading this raspberry pi

my raspberry pi still has the memories of the last time i used it (2015, rsyncing)

not that 1&2 are bad or inferior to consider. those sources of variability capture different information - would I have gotten a different model if training had been different?

Show thread

I think ML researchers are getting more into reporting estimates of variance of things, but the *origin* of the variance receives less attention. is it due to...?
1. stochasticity of optimisation
2. variation in training dataset
3. variation in test dataset

1&2 are about the process of buiding the model
3 is about the actual generalisation performance

Show thread

gotten weirdly into bootstrapping. I'll bootstrap anything, but preferably a confidence interval for test set performance

I love exploring new datasets and googling all the stuff. Today I learned about Foley catheters and IV pushes and the fact that the same type of infusion can be recorded with multiple different units (in this one dataset).

Meanwhile another 80 machine learning papers appear on arXiv

hello everyone, I'm stephanie (she/her)!

I'm a researcher at a big tech company, based in the UK. I did my phd in computational biology+medicine. I focus very much on ML for health, specifically critical care, physiological time series, "how to do ML well in health", etc.

Non-research activities include rollerskating/derby, gaming (computer/table-top/board), sewing, looking at birds.

Scholar Social

Scholar Social is a microblogging platform for researchers, grad students, librarians, archivists, undergrads, academically inclined high schoolers, educators of all levels, journal editors, research assistants, professors, administrators—anyone involved in academia who is willing to engage with others respectfully.