Top
    Welcome to SLR Toolbox Web Page 
What is SLR toolbox?
The Sparse Logistic Regression toolbox (SLR toolbox hereafter) is a suite of MATLAB functions for solving classification problems. It provides one of solutions for binary or multi-class classification problem. The unique feature is weight parameters of the classifier are learned in a sparse way, implementing automatic feature selection during learning weight parameters.
History
  • 2016 March 31th
    Release ver 1.51. Iterative SLR for binary classification is implemented. Fix a bug of error table creation when only one single label appears in test or train data.

  • 2015 Feb.25th
    To resolve the over-sparse problem of SLR, Dr.Hirose et al. in CiNET recently proposed the iSLR method in which SLR is applied iteratively and sequentially to features unselected by SLR in the previous steps. You can download the MATLAB code from his homepage.

  • 2009 Aug.10th: Version 1.21alpha
    Minor bugs fixed.

  • 2009 July 28th : Version 1.2 alpha
    Two new binary calssfication algortihms (L1-norm based SLR) are implmented (but has not been tested carefully).

  • 2009 June 5th Version 1.1alpha
    A new multiclass classifier (SLR based one-versus-one classifier) is added.

  • 2009 June 3rd : Version 1.0beta
  • Features
  • Sparse parameter estimation
    • Selection of relevant features during estimating weight parameters of a classifier
    • Appropriate for classification problems with high dimensional features
    • Avoid overfitting to some extent
  • Basically no need to tune parameters in algorithms
  • Covered Algorithms
  • Four types of binary classifiers (seven algorithms)
  • Four types of multi-class classifiers (six algorithms)
  • Note : In this toolbox, several variants of classifiers including logistic regression(LR), ARD-SLR, L1-regurlarized SLR, L2-regularized LR and relevance vector machine are implemented. However the ARD-based SLR and SMLR(multi-class version) will be mainly maintained and supported in the future.

    Environment
    The codes in the toolbox were written for MATLAB ver7.0.1 or later under UNIX.
    Several functions require the optimization toolbox.
    Download
  • SLR toolbox ver1.51(126kb)
  • SLR toolbox ver1.21alpha(122kb)
  • SLR toolbox ver1.2alpha(148kb)
  • SLR toolbox ver1.1alpha(108kb)
  • SLR toolbox ver1.0beta (102kb)
  • Test data (2290kb)
  • Documentation
  • Readme Document (.pdf)
  • : Please read this document at first to understand contents of the toolbox.
  • History of SLR toolbox (pdf)
  • : Release note
  • Mathematical Issues(pdf)
  • : This document covers the mathematical background of the implemented algorithms (for binary classifiers).
    Installation
    To get installed the toolbox, you just download and unzip the file ('SLR*.zip') wherever you like. You may also download test-data sets that were acquired from two real experiments. Please unzip 'TESTDATA.zip' in the same level as 'SLR*'. They are optionaly used in two demo functions ('demo_binary_classification.m' and 'demo_multiclass_classification.m'). Please start from demo functions ('demo_*.m') to learn how the functions in SLR toolbox work.
    References

    For the applications of this method to the fMRI decoding studies, please see the following papers.

    Yamashita O, Sato MA, Yoshioka T, Tong F, Kamitani Y (2008).
    Sparse estimation automatically selects voxels relevant for the decoding of fMRI activity patterns. Neuroimage. Oct 1;42(4):1414-29

    Miyawaki Y, Uchida H, Yamashita O, Sato MA, Morito Y, Tanabe HC, Sadato N, Kamitani Y (2008).
    Visual image reconstruction from human brain activity using a combination of multiscale local image decoders. Neuron. Dec 10;60(5):915-29.

    Copyright
    SLR toolbox is free but copyright software, distributed under the terms of the GNU General Public Licence as published by the Free Software Foundation. Further details on "copyleft" can be found at http://www.gnu.org/copyleft/. No formal support or maintenance is provided or implied. SLR BSD license version is also available from here.
    Feedback

    oyamashi@atr.jp