Administrative tasks can lead patients to delay or forgo health care, but Michael Anne Kyle thinks there are ways to change that. https://t.co/vgT5h3MTjk
@tcarpenter216 If no result, or if you think your data covers all possible categorical values, stratified is okay. If your model gives some output when faced with new cat. levels, then stratified may lead to overfitting
@tcarpenter216 A few questions: does your data cover all possible cat. values? In your final use case, could there be a category value outside your train+test data. How is your model geared to deal with that, no result, or something based on some interpolation
Why is competence in transitioning between levels of abstraction so important? Determining what's essential and what's not is a critical skill for any researcher. By learning to selectively focus on certain aspects of reality and ignore others, students come to understand that what gets into a model is the decision of the researcher, not a statistical analyst or fancy algorithms.
This is a good read
Clinicians keep being asked to do more with less, or forced to work with tools and in conditions they repeatedly report as being unsafe, and then when things go wrong, they are often the ones blamed
https://t.co/VGqxVMcj3x
I like this framework to evaluate where ChatGPT would be useful. I got ChatGPT to shorten one of my abstracts to within the word limit for a conference. Tweaking that usually takes me some time, but verifying correctness is easy (it's my research!)
Generative models and P vs. NP: A clickbaity thread 🧵
An important point that seems missing (as far as I've seen) in the debate over LLM "correctness" / "usefulness": generative models are most useful precisely in situations where creation is hard, but verification is easy. 1/N
What connects @rashtrapatibhvn Droupadi Murmu, @PrimeVideoIN show Panchayat & @NobelPrize winning Economist Esther Duflo?
The 73rd amendment of Constitution of India that empowered local governments and introduced women’s reservation in local governments.
A Thread 🧵
So I'm working on a little system to create a personal academic webpage from a google sheet automatically. It takes a public google sheet as input (https://t.co/qRxqHXeken) and makes a GitHub page (https://t.co/Pz7zVc2OT0) automatically
"[John von Nuemann] always contended with Dantzig that the simplex method would take an absurdly long amount of time to solve linear programming problems. It appears to be, oh, so far as I know, the one place where Johnny went very badly wrong." -- Philip Wolfe