Mikio L. Braun

ML and AI expert

Back to archive

Thread

3 tweets

1
Yeah, probably not :) One of my pet peeves about scikit-learn‘s design is that you instantiate a model and the call its fit() function that turns it into something you can call predict() on. It always seemed much cleaner to me to have a Learner whose fit() would product a Model
2
2
@dirkriehle that has predict(). As it is, it is (a) too stateful() and also (b) not cleanly designed. Sorry, not exactly your question, but related. My guess is that the MLOps people are transferring DevOps terms, but sometimes there is no clear analogy.
1
3
@dirkriehle My other pet peeve is that MLOps people focus too much on the finished pipeline and less on the process to find it. That would probably also lead to a different set of analogies and terms. To me, it‘s more like a workbench in a chemistry lab.
1