About the Machine Learning category

Machine Learning in Julia, with a particular focus on Deep Learning

  • Flux
  • Knet
  • Tensorflow.jl
  • Differentiable programming
2 Likes

Some MLJ developers and myself are inclined to move more of our current slack discussions to Discourse to make them archival and more discoverable. However, I wonder however if the “Machine Learning” category is best suited to our purpose.

Despite the name, discussions (and the packages listed under “About the Machine Learning category”) are 99% “deep learning”, whereas our users are more interested in multi-paradigm machine learning. (MLJ has over 160 models, and maybe half a dozen neural network models.) The same would apply to the users of AutoMLPipeline (cc @ppalmes) and ScikitLearn (@cstjean). Given the reasonable level of traffic in this forum, I wonder if the “Machine Learning” category is a good match for our needs.

Maybe the community would be open to one of these???

  1. rename “Machine Learning” → “Deep Learning” and add “Machine Learning (general)” ; or
  2. add a “Machine Learning Toolboxes” category (for MLJ, ScikitLearn, AutoMLPipeline, BetaML, …)

Option 1 sounds better to me.

Thoughts anyone?

@vchuravy @tlienart @samuel_okon @vollmersj @dilumaluthge @juliohm @cscherrer

9 Likes

i also agree that since deep learning or any gradient-based approach is just a subset of the broader machine learning category, it will be good if we can make this difference explicit to avoid confusion and people can ask their questions to the right discussion group.

My personal preference is to leave the topic Machine Learning untouched and just interpret it as the usual broader area of research that includes “deep” learning as a subtopic. I have a hard time myself accepting “deep” learning as a separate thing.

BTW, thank you for considering forums with archival. It is a very good move. You could also consider moving from Slack to Zulip for instantaneous support + community building. Slack is a disservice to the Julia community.

4 Likes

Also note that the description above is a wiki and editable by anyone with TL2 (I think?).
So if you find the description to narrow please extend it! We can also add specific tags for specific tools.