Skip to content

Welcome to sawmil's documentation!

This package is in the alpha stage of testing.

sAwMIL (Sparse Aware Multiple-Instance Learning) is an open-source Python library providing a collection of Support Vector Machine (SVM) classifiers for multiple-instance learning (MIL). It builds upon ideas from the earlier misvm package, adapting it for the latest Python version, as well as introducing new models.

In Single-Instance Learning (SIL), the dataset consists of pairs of an instance and a label:

\[ \langle \boldsymbol{x}_i, y_i \rangle \text{ , where } \boldsymbol{x}_i \in \mathbb{R}^{d} \text{ and } y_i \in \mathcal{Y}. \]

In binary settings, the label is \( y \in \{0,1\} \). To solve this problem, we can use a standard SVM model.

In Multiple-Instance Learning (MIL), the dataset consists of bags of instances paired with a single bag-level label:

\[ \langle \boldsymbol{X}_i, y_i \rangle \text{ , where } \boldsymbol{X}_i = \{ \boldsymbol{x}_{1}, \boldsymbol{x}_{2}, ..., \boldsymbol{x}_{n_i} \}, \boldsymbol{x}_j \in \mathbb{R}^{d} \text{ and } y_i \in \mathcal{Y}. \]

To solve this problem, we can use NSK or sMIL models.

In some cases, each bag, along with the instances and a label, could contain a intra-bag mask that specifies which items are likely to contain the signal related to \(y\). In that case, we have a triplet of \( \langle \boldsymbol{X}_i, \boldsymbol{M}_i, y_i \rangle \), where

\[ \boldsymbol{M}_i = \{m_1, m_1,... m_{n_i}\}, \text{ where } m_j \in \{0,1\}. \]

To solve this problem, one can use the sAwMIL model.

Solvers

Our package supports three QP backends: Gurobi, OSQP and DAQP.

Models

Single-Instance SVMs

Our package implements a simple SVM for the SIL setting.

Multiple-Instance SVMs

The multiple-instance SVMs are adapted to work with the bags

  1. NSK (Normalized Set Kernel) 1
  2. sMIL (Sparse Multiple-Instance Learnig) 2
  3. sAwMIL (Sparse Aware MIL) 3 is used in trilemma-of-truth.

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

  2. Razvan C Bunescu and Raymond J Mooney. Multiple instance learning for sparse positive bags. In Proceedings of the 24th international conference on Machine learning, 105–112. 2007. ↩

  3. Germans Savcisens and Tina Eliassi-Rad. The trilemma of truth in large language models. arXiv preprint arXiv:2506.23921, 2025. ↩