This package provides Python code for SamBA, a boosting-based classifier that uses similarity to learn a combination of local experts. For in-depth information, see B.Bauvin et al, UAI 2023.
It heavily relies on scikit-learn, for usage examples, see the demo directory in which the code used for the paper's figures is provided.
Once you cloned the project, run, with python3:
cd path/to/neighborhood_classifier/
pip install -e .
To reproduce the results, we provide the demo directory, in which each
experiment is coded. They are straightforward, except for the performance study
that heavily relies on SuMMIT.
We are aware that the tool is too complex for the simple experiments provided
in the paper. As a consequence, we provided a minimal version of summit in the
Performance directory.
To reproduce the results of the paper run:
cd path/to/neighborhood_classifier/demo/summit_minimal_version/
pip install -e .
cd summit/
python execute --config_path ../../config_summit_metagenome.yml
Such an execution might be long and demanding. As a consequence, we provide a
pre-computed result directory Performance/results/metagenome/pre_computed/,
that is used in the Performance/preformance_study.py script.
The metagenome dataset is provided in the Performance/dataset/ directory
as an HDF5 file containing all the data types of the paper.
In addition, we provide a csv version for each data type, and one for the labels.
@InProceedings{pmlr-v216-bauvin23a,
title = {Sample {B}oosting {A}lgorithm ({SamBA}) - An interpretable greedy ensemble classifier based on local expertise for fat data},
author = {Bauvin, Baptiste and Capponi, C\'{e}cile and Clerc, Florence and Germain, Pascal and Ko\c{c}o, Sokol and Corbeil, Jacques},
booktitle = {Proceedings of the Thirty-Ninth Conference on Uncertainty in Artificial Intelligence},
pages = {130--140},
year = {2023},
editor = {Evans, Robin J. and Shpitser, Ilya},
volume = {216},
series = {Proceedings of Machine Learning Research},
month = {31 Jul--04 Aug},
publisher = {PMLR},
pdf = {https://proceedings.mlr.press/v216/bauvin23a/bauvin23a.pdf},
url = {https://proceedings.mlr.press/v216/bauvin23a.html},
abstract = {Ensemble methods are a very diverse family of algorithms with a wide range of applications. One of the most commonly used is boosting, with the prominent Adaboost. Adaboost relies on greedily learning base classifiers that rectify the error from previous iterations. Then, it combines them through a weighted majority vote, based on their quality on the entire learning set. In this paper, we propose a supervised binary classification framework that propagates the local knowledge acquired during the boosting iterations to the prediction function. Based on this general framework, we introduce SamBA, an interpretable greedy ensemble method designed for fat datasets, with a large number of dimensions and a small number of samples. SamBA learns local classifiers and combines them, using a similarity function, to optimize its efficiency in data extraction. We provide a theoretical analysis of SamBA, yielding convergence and generalization guarantees. In addition, we highlight SamBA’s empirical behavior in an extensive experimental analysis on both real biological and generated datasets, comparing it to state-of-the-art ensemble methods and similarity-based approaches.}
}
- Baptiste BAUVIN