a python 2.7 implementation of lagrangean-s3vm
*the core routine is in lagrangian_s3vm.py; see method documentation for details
*utils.py contains:
- a cross-validation utility method (basically wrapping that of sklearn); lagrangean-s3vm uses it for initializing a supervised model on the labeled training set
- a method which loads the COIL20 dataset
*datasets/ contains a convenient dump (a few MBs) of the COIL20 dataset, which can be used for testing
*test.ipynb is a jupyter notebook, showing how to use each method on a COIL20 classification instance
#####################################
please cite us if you use this software!
@article{Bagattini18_a, title={A Simple and Effective Lagrangean-based Combinatorial Algorithm for {S3VM}s}, author={F. Bagattini and P.Cappanera and F. Schoen}, journal={Proceedings of the Third International Conference on Machine Learning, Optimization and Big Data}, year={forthcoming 2018}, publisher={Springer} }
@article{Bagattini18_b, title={Lagrangean-based Combinatorial Optimization for Large Scale {S3VM}s}, author={F. Bagattini and P.Cappanera and F. Schoen}, journal={IEEE Transactions of Neural Networks and Learning Systems}, year={forthcoming 2018}, publisher={IEEE} }