SlipGURU Dipartimento di Informatica e Scienze dell'Informazione Università Degli Studi di Genova

L1L2Py Documentation

Release:1.0.5
Date:January 25, 2012
Homepage:http://slipguru.disi.unige.it/Software/L1L2Py
Download:http://slipguru.disi.unige.it/Software/L1L2Py/L1L2Py-1.0.5.tar.gz

L1L2Py is a Python package to perform feature selection by means of \ell_1\ell_2 regularization with double optimization.

L1L2Py makes use of NumPy to provide fast N-dimensional array manipulation. It is licensed under GNU General Public License (GPL) version 3.

L1L2Py is based on the minimization of the (naive) \ell_1\ell_2 functional introduced in [Zou05] using the algorithm studied from the theoretical viewpoint in [DeMol09a]. The current implementation exploits the minimization algorithm proposed in [Beck09].

L1L2Py is the Python implementation of the one proposed and applied in [DeMol09b]. It consists of two stages. The first one identifies the minimal set of relevant variables (in terms of prediction error). Starting from the minimal list, the second stage selects the family of (almost completely) nested lists of relevant variables for increasing values of linear correlation.

Index

References

[Zou05]H. Zou, T. Hastie, “Regularization and variable selection via the elastic net” J.R. Statist. Soc. B, 67 (2) pp. 301-320, 2005
[DeMol09a]C. De Mol, E. De Vito, L. Rosasco, “Elastic-net regularization in learning theory” Journal of Complexity, n. 2, vol. 25, pp. 201-230, 2009.
[DeMol09b]C. De Mol, S. Mosci, M. Traskine, A. Verri, “A Regularized Method for Selecting Nested Group of Genes from Microarray Data” Journal of Computational Biology, vol. 16, pp. 677-690, 2009.
[Beck09]A. Beck, M. Teboulle, “A fast iterative shrinkage-thresholding algorithm for linear inverse problems” SIAM Journal on Imaging Sciences, 2(1):183–202, Mar 2009.

Download

Current version: 1.0.5

Get L1L2Py from the Python Package Index, or install it with:

pip install -U L1L2Py
or:
easy_install -U L1L2Py

Latest documentation in pdf is also available.