The friendly competition between R and python has gifted us with two stellar packages for workflow-style predictive modeling: tidymodels in R, and scikit- learn in python. When I had to choose between them for a Machine Learning Course, I said: ¿Porque no los dos? (Why not both?)
In this talk, I will share how the differences in structure and syntax between tidymodels and scikit-learn impacted student understanding. Can a helper function hide an important decision about tuning parameters? Can a slight change in argument input influence the way we describe a model? The answer is a resounding, "¡Sí!"
Don't despair, though, because I will also provide advice for avoiding pitfalls when switching between languages or implementations. Together, let's think about the power that programming choices has to shape the mental model of the user, and the ways that we can responsibly document our modeling decisions to increase cross-language reproducibility.
Talk materials are available at https://www.kelly-bodwin.com/talks/rsconf22/.
Kelly Bodwin is an Assistant Professor of Statistics and Data Science and an RStudio Certified Instructor. She is passionate about developing R tools (especially choosing names for packages) and about creating innovative ways to teach computing to data scientists. Although her love for R is eternal and unwavering, she is currently begrudgingly allowing a little python into her heart.
