Multi-Objective Simultaneous Model and Feature Selection

Performs simultaneous hyperparameter tuning and feature selection through both single-objective and multi-objective optimization as described in Binder, Moosbauer et al. (2019) . Uses the 'ecr'-package as basis but adds mixed integer evolutionary strategies and multi-fidelity functionality as well as operators specific for the problem of feature selection.

MOSMaFS: Multi-Objective Simultaneous Model and Feature Selection

Project Status

We are on CRAN! You can also read our paper on arxiv.

Installation

It is best to install the official version from CRAN. If you want to live on the edge, you can also install the github version, using remotes:

remotes::install_github("slds-lmu/mosmafs", ref = "mosmafs-package")

Documentation

Citation

If mosmafs is useful for your research, please cite our paper:

Binder, Martin., Moosbauer, J. et al. "Model-Agnostic Approaches to Multi-Objective Simultaneous Hyperparameter Tuning and Feature Selection." arXiv preprint arXiv:1912.12912 (2019).

BibTex:

@misc{binder2019multiobjective,
    title={Multi-Objective Hyperparameter Tuning and Feature Selection using Filter Ensembles},
    author={Martin Binder and Julia Moosbauer and Janek Thomas and Bernd Bischl},
    year={2019},
    eprint={1912.12912},
    archivePrefix={arXiv},
    primaryClass={stat.ML}
}