Skip to content

Welcome!

This package is in the alpha stage of testings.

sAwMIL provides Multiple-Instance Learning (MIL) models built on support vector machines. It is inspired by the outdated misvm package. Our package supports two QP backends: Gurobi and OSQP.

Single-Instance SVMs

  1. SVM (Single-Instance SVM)

Multiple-Instance SVMs

  1. NSK (Normalized Set Kernel) 1 See the exapliner on NSK class for details.
  2. sMIL (Sparse MIL)

    Bunescu, Razvan C., and Raymond J. Mooney. Multiple instance learning for sparse positive bags. Proceedings of the 24th International Conference on Machine Learning (2007).

  3. sAwMIL (Sparse-Aware MIL; two-stage: sMIL → instance SVM) Classifier used in trilemma-of-truth: > Savcisens, Germans, and Tina Eliassi-Rad. The Trilemma of Truth in Large Language Models. arXiv preprint arXiv:2506.23921 (2025).

  1. Thomas Gärtner, Peter A Flach, Adam Kowalczyk, and Alexander J Smola. Multi-instance kernels. In ICML, volume 2, 7. 2002.