Introduction to Machine Learning with {tidymodels}

preview_player
Показать описание
Workshop recorded as part of the R/Pharma Workshop Series (October 18, 2023)
Instructors: Nicola Rennie (Lancaster University)

Resources mentioned in the workshop:
Рекомендации по теме
Комментарии
Автор

1:03:36 I think the difference is that the models being fit during hyperparameter tuning are fitted using (k-1)/k part of the training data. That’s because the last part is used for evaluation during the k-fold tuning. So when we know what the best hyperparameters are we want to fit the model using the whole training set.

JerryWho
Автор

38:59 Shouldn’t it be the sum of the **absolute values** of the coefficients?

JerryWho
Автор

This is a very well constructed lecture and reproducible (all the data and source code is given on github). A nice scaffold to build your own lectures on and expand. My only complaint is that the transitions are often not clear and bit rugged (e.g. spend a lot time on LASSO but then use Logistic Regression).

haraldurkarlsson
Автор

If you are worried about which columns are picked in step_normalize() and you all want columns with values greater than 1 then I believe this code works: &&
any(.x > 1))). Now what the author uses in the video is more straight forward and thus simpler but if you have a lot of columns the first approach might be safer.

haraldurkarlsson
Автор

If you are starting out with tidymodels then yo might be confused since a little of details are left out. Naturally you cannot cover such a big subject in a short lecture. Those needing more information might want to look at "Tidy Modeling with r" by Kuhn and Silge(2022). A free ebook version is available online.

haraldurkarlsson
join shbcf.ru