Another excellent lecture! Your content has been helping a lot to improve my modelling and R skills. To be honest, I almost completely migrated to tidymodels as my main framework to modelling. There are so many handful tools in tidymodels for cross validation, tuning, feature engineering, etc... that it is hard to go back when you get used to it. The only reason why I didn't fully migrated is due to keras integration of multilayer nn, which I have to use a lot and I feel like it is not there yet. Anyways, any chance of a lecture on integrating keras and tidymodels any time soon? Thx for the amazing work.
You might be interested in the last third of SMLTAR, where we demonstrate how to use tidymodels packages together with keras: smltar.com/dloverview.html
@@JuliaSilge wow! Thx so much Julia, I was already using some tidymodels functionalities from yardstick to get more detailed metrics from keras, etc. However, I was really struggling to do cross validation once most examples are based on for loops. The text you sent just adressed my struggles!
Hello Julia, thank you for your so interesting articles. I'm a bit puzzled as I recently saw a last one that I try to find over but am not able. It is about creating and publishing an API using plumber, pins, butcher packages... I'm getting crazy causevI'm sure I saw it!! Is there a reason why it doesn't show anywhere anymore? Thanks again, Laurent
Ah, it is over here at the RStudio RU-vid: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-3VJ5cIxe408.html You may also be interested in our new package vetiver: vetiver.tidymodels.org/
Hey Julia - quick question on a random topic. when using fit_resamples and save predictions on a classification data set, why does it not save the probabilities only the class prediction? I used to be able to create a roc curve on all the ids of each resample, but now I'm only getting the class prediction itself and not probabilities to create a roc curve. Thanks for any help!
I"m guessing either the model doesn't support probabilities or you set the metric set or something like that. If you can create a reprex demonstrating what happens and post it on RStudio Community: rstd.io/tidymodels-community folks will be able to take a look.
