Scikit-mlm

A first full-stable version.

new features:
- fix bugs on L12_MLM

News:
- imlementation of [ℓ_1/2-norm regularization MLM (L12_MLM)](https://doi.org/10.1109/BRACIS.2018.00043) technique.

implemented methods
original proposal
- [regression (MLMR)](https://doi.org/10.1016/j.neucom.2014.11.073)
- [classification (MLMC)](https://doi.org/10.1016/j.neucom.2014.11.073)

speed up
- [nearest neighbor MLM (NN_MLM)](https://link.springer.com/article/10.1007%2Fs11063-017-9587-5Sec9)
<!-- - [ ] [cubic equation MLM (C-MLM)](https://link.springer.com/article/10.1007%2Fs11063-017-9587-5Sec10) -->

reference points selection methods
classification
- [opposite neighborhood MLM (ON_MLM)](https://www.elen.ucl.ac.be/Proceedings/esann/esannpdf/es2018-198.pdf)
- [fuzzy C-means MLM (FCM-MLM)](https://doi.org/10.1007/978-3-319-95312-0_34)
- [optimally selected MLM (OS-MLM)](https://doi.org/10.1007/978-3-030-03493-1_70)
<!-- - [ ] [&ell;_1/2-norm regularization MLM (L12_MLM)](https://doi.org/10.1109/BRACIS.2018.00043)
-->

<!-- regression
- [ ] [regularized M-FOCUSS MLM (RMF_MLM)]() -->

cost Sensitive
- [weighted MLM (w_MLM)](https://doi.org/10.1007/978-3-319-26532-2_61)