Mill.jl (Multiple Instance Learning Library) is a library aimed to build flexible hierarchical multi-instance learning models built on top of Flux.jl. It is developed to be:

• flexible and versatile
• as general as possible
• fast
• and dependent on only handful of other packages

Watch our introductory talk from JuliaCon 2021

Run the following in REPL:

] add Mill

Julia v1.9 or later is required.

• Documentation
• API Reference
• Examples

Kindly cite our work with the following entries if you find it interesting, please:

• JsonGrinder.jl: automated differentiable neural architecture for embedding arbitrary JSON data

  @article{Mandlik2022,
   author = {Šimon Mandlík and Matěj Račinský and Viliam Lisý and Tomáš Pevný},
   issn = {1533-7928},
   issue = {298},
   journal = {Journal of Machine Learning Research},
   pages = {1-5},
   title = {JsonGrinder.jl: automated differentiable neural architecture for embedding arbitrary JSON data},
   volume = {23},
   url = {http://jmlr.org/papers/v23/21-0174.html},
   year = {2022},
  }
  

• Malicious Internet Entity Detection Using Local Graph Inference (practical Mill.jl application)

  @article{Mandlik2024,
    author  = {Mandlík, Šimon and Pevný, Tomáš and Šmídl, Václav and Bajer, Lukáš},
    journal = {IEEE Transactions on Information Forensics and Security},
    title   = {Malicious Internet Entity Detection Using Local Graph Inference},
    year    = {2024},
    volume  = {19},
    pages   = {3554-3566},
    doi     = {10.1109/TIFS.2024.3360867}
  }
  

• this implementation (fill in the used version)

  @software{Mill,
    author  = {Tomas Pevny and Simon Mandlik},
    title   = {Mill.jl framework: a flexible library for (hierarchical) multi-instance learning},
    url     = {https://github.com/CTUAvastLab/Mill.jl},
    version = {...},
  }
  

If you want to contribute to Mill.jl, be sure to review the contribution guidelines.

We use GitHub issues for tracking requests and bugs.

Icon source