EPFD

Code of Paper:
Ensemble Pruning based on Objection Maximization with a General Distributed Framework

Including:

• Centralized Objection Maximization for Ensemble Pruning (COMEP)
• Distributed Objection Maximization for Ensemble Pruning (DOMEP)
• Ensemble Pruning Framework in a Distributed Setting (EPFD)

Dependencies

• Create a virtual environment

  $ conda create -n EPFD python=3.9
  $ source activate EPFD
  $ # source deactivate

  or

  $ virtualenv EPFD --python=/usr/bin/python3
  $ source EPFD/bin/activate
  $ # deactivate

• Install packages

  $ pip install -r requirements.txt
  $ git clone https://github.com/eustomaqua/PyEnsemble.git
  $ pip install -e ./PyEnsemble

Examples

Dataset: iris

Optional Choices of Ensemble Pruning Methods:
name-pru ['ES', 'KP', 'KL', 'RE', 'OO', 'DREP', 'SEP', 'OEP', 'PEP', 'COMEP', 'DOMEP']

e.g.,

$ python main.py --nb-cls 31 --nb-pru 7 --name-pru COMEP --lam 0.5 --m 2
$ python main.py --nb-cls 31 --nb-pru 7 --name-pru DOMEP --lam 0.5 --m 2
$ python main.py --nb-cls 31 --nb-pru 7 --name-pru PEP --distributed --m 2

Cite

Please cite our paper if you use this repository

@article{8891828,
  title     = {Ensemble Pruning Based on Objection Maximization With a General Distributed Framework}, 
  author    = {Bian, Yijun and Wang, Yijun and Yao, Yaqiang and Chen, Huanhuan},
  journal   = {IEEE Transactions on Neural Networks and Learning Systems}, 
  year      = {2020},
  volume    = {31},
  number    = {9},
  pages     = {3766--3774},
  doi       = {10.1109/TNNLS.2019.2945116},
  publisher = {IEEE},
  url       = {https://ieeexplore.ieee.org/document/8891828},
}