MLJ (Machine Learning in Julia) 0.12 update


Try it out (nothing to install): Binder

Docs: MLJ

Learn: Data Science Tutorials in Julia.

Free: workshop at JuliaCon2020

Been a little while since our v0.1 announcement. Now at v0.11.6 v0.12.0, with functionality spread over over ten repositories:

MLJ has been under continuous development since November 2018. Worth a try if you have not looked at it for a while.

MLJ is a toolbox written in Julia providing a common interface and meta-algorithms for selecting, tuning, evaluating, composing and comparing machine learning models written in Julia and other languages. Now with access to over 140 models, including most popular scikit-learn models, as well as deep learning models via MLJFlux, recently released.

MLJ strengths include:

  • Flexible, carefully thought out API for model composition, including new improved @pipeline and @from_network syntax

  • Extendible hyper-parameter optimization interface that plays well with model composition, with multiple resampling and (nested) parallelization options. Recently added algorithms are:

    • RandomSearch with customizable priors (built-in)

    • MLJTreeParzenTuning provided by TreeParzen.jl

  • Searchable database of model metadata for matching models to learning tasks

Developers should note that MLJ has extensive documentation for integrating a new machine learning model into MLJ and for adding a hyper-optimization algorithm.


I appear to have v0.12.0 on my machine. :slight_smile:

Can you quickly zip that and send it to me. Save me a lot of work!! :laughing:

1 Like

I’m confused. It looks like v0.12 came before 11.5:

Was there a mixup in version naming or is the point of v0.11.6 to continue to make v0.11 bugfix releases while simultaneously moving forward with breaking releases?

1 Like

Okay egg on my face. I made a post 0.12 fix to 0.11.5 that I had completely forgotten about!!

Current version is indeed 0.12.0