Robust Minimax Boosting with Performance Guarantees (RMBoost)

This repository is the official implementation of Robust Minimax Boosting with Performance Guarantees

RMBoost methods are robust to general types of label noise and can also achieve strong classification performance.

Source code

mMBoost folder contains the Python and Matlab folders that include the Python and Matlab implementations, respectively.

Python code

• run_RMBoost.py is the main file. In such file we can modify the number of rounds and the solver (linprog or mosek)
• RMBoost.py is the file that includes fit and predict functions

Requirements

The requirements are detailed in the requeriments.txt file. Run the following command to install the requeriments:

pip install -r requirements.txt

Matlab code

• main.m is the main file. In such file we can modify the number of rounds and the solver (linprog or mosek)
• fit.m is the function that fits the model
• predict_boost.m is the function that obtains the predictions

Installation and evaluation

To train and evaluate the model in the paper, run this command for Python:

python run_RMboost.py

and for Matlab:

matlab RMBoost.m

Support and Author

Santiago Mazuelas

smazuelas@bcamath.org

Verónica Álvarez

vealvar@mit.edu

License

RMBoost carries a MIT license.

Citation

If you find useful the code in your research, please include explicit mention of our work in your publication with the following corresponding entry in your bibliography:

@inproceedings{MazAlv:25, title ={Robust Minimax Boosting with Performance Guarantees}, author ={Mazuelas, Santiago and {'A}lvarez, Ver{'o}nica}, booktitle ={{A}dvances in {N}eural {I}nformation {P}rocessing {S}ystems}, volume ={38}, pages ={}, year ={2025}, month ={Dec.} }