Learning with Partial Supervision
Project description
- Topic:
Generic implementations of weakly supervision algorithms developped in [CAB20], [CAB21a], [CAB21b].
- Version:
- 1.0.0 of 2021/06/07
Installation
To install our package, run the setup file
$ python <path to code folder>/setup.py install
You can also install it in develop mode, eventually with pip
$ cd <path to code folder>
$ pip install -e .
Usage
- See files:
problems/classification/libsvm_experiments.py
problems/classification/semi_supervision_experiments.py
and more generally *_experiements.py
Package Requirements
- Most of the code is based on the following python libraries:
numpy
numba
matplotlib
- Some testing done with notebook are based on:
jupyter-notebook
ipywidgets
- For ranking, we used the following lp solver library:
cplex
- To load LIBSVM files, more precisely to read libsvm files format we used:
scikit-learn
- To load MULAN files, more precisely to read mulan files format we used:
arff
skmultilearn
Datasets links
- Datasets can be download at:
Change path in config file dataloader/config.py to specify path to your data.
References
Structured Prediction with Partial Labelling through the Infimum Loss, Cabannes et al., ICML, 2020
Disambiguation of weak supervision with exponential convergence rates, Cabannes et al., ICML, 2021
Overcoming the curse of dimensionality with Laplacian regularization in semi-supervised learning, Cabannes et al., Preprint, 2021
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.