Zhiyong Yang, Qianqian Xu, Shilong Bao, Yuan He, Xiaochun Cao and Qingming Huang. Learning with Multiclass AUC: Theory and Algorithms. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2021. (Regular Paper)
Note that, this repo only supplies the code of AUC optimization for traditional ML methods. We have open-sourced an end2end machine learning library for X metrics learning (including AUROC for deep neural networks). Please refer to XCurve for the official code. We hope our library can help you deploy/attain your ML model conveniently and easily. Thanks for all!
This implementation is inspired by scikit-learn, and we extended it to our work (especially, the special cv regime).
- numpy
- scikit-learn
- pandas
- pickle
Cython is an efficient extension for python to accelerate the code.
We also accelerate this code with cython and achieve remarable performance!(Refer to cfunc.pyx and setup.py)
To execute experiments with cython, u should compile the cython code first .
python3 setup.py build_ext --inplace
cython -a cfunc.pyx
Then, set the bool cython=True, to enjoy the acceleration.
When u have already set the corresponding parameters, then
python3 train.py