Kelly Bodwin | Translating from {tidymodels} and scikit-learn: Lessons from a 'bilingual' course

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 www.kelly-bodw...
Session: Teaching data science

[@jimcallahan448]

Great talk. Some parts of R and Python are very similar such as read_csv() in the tidyverse and pandas where both produce a data frame. But, in Python the adoption of the data frame is limited and as one progresses in machine learning one is dealing with numpy arrays and tensors rather than data frames. Good example in "cost" vs. "c" parameters. I tend to trust R more but like Python's spirit of experimentation.