Do people no longer utilize classic libraries like Sci-Kit Learn?

Hot Discussions Forum
1

I just came across a tweet, which gained a lot of attention, discussing how many data scientists haven’t used Sci-Kit Learn in months.

PyTorch, HuggingFace, langchain, supergradients, etc. have taken its place.

I’m not sure if I’m just behind the times or what you guys think/do, but I’m wondering and a little concerned that I might be. This didn’t really make sense to me because the tooling stated isn’t really equivalent to sci-kit learn.

1 Like
2

The majority of data science problems can and should be tackled at least partially with traditional statistics. Saying stuff like “pytorch and huggingface have made sklearn obsolete” is a fantastic way to expose yourself as a bullshit artist, I suppose. To be honest, this represents a sizable portion of the field.

1 Like
3

You’re not falling behind since those are distinct instruments. For standard machine learning, Sci-Kit Learn is still extensively used; for deep learning and natural language processing, PyTorch, HuggingFace, and other programs are well-liked. Use what is appropriate for your project.

1 Like
4

Yes, people still use classic libraries like Sci-Kit Learn a lot. Here’s why:

  • Most data people work with is in tables, and traditional machine learning models are best for this.
  • Sci-Kit Learn is still excellent for training models on tabular data.
  • Newer tools like PyTorch and deep learning models are often not suitable for tabular data.
1 Like