HCclassification

This archive contains a Matlab implementation of HC-Classification algorithm as described in the paper 
D. Donoho and J. Jin (2008) Higher criticism thresholding: Optimal feature selection when useful features 
are rare and weak


Current version 

V1.1


SHORT DOCUMENTATION

For more information, type help functionname in the Matlab command line.

There are two functions in this package, HCclassification to find the classifier, and HCclassification_fit 
to apply the classifier to new data. 

Usage: 
    [weight, stats] = HCclassification(TrainX, TrainY, threshold, alpha, sdflag, muflag)

Inputs:
  TrainX   N-by-P matrix of predictors for train data set with one row per observation and one column per predictor.
  TrainY   N-by-1 matrix of class labels "1" or "0"

Inputs (Optional):
  threshold Choice of thresholding functions, can be 'clip', 'soft', or 'hard'
           default 'hard'
  alpha    The proportion of features with small p-values to calculate HC threshold. 
           default  0.2
  sdflag   optional type of proxy sd: 
             0 - std; 2 - floored at median; 1 - add median ;
           default  1
  muflag   optional type of proxy mu: 0, std, 1, average two means
           default 1

Outputs
  weight   p-by-1 vector, showing the weight for each feature
  stats    4-by-1 struct, including the normalization parameters for test data and the statistic
     stats.xbar   p-by-1 chosen mu proxy
     stats.s      p-by-1 chosen standard deviation proxy
     stats.HCT    chosen threshold
     stats.HC     p-by-1 HC score for each feature



Usage:
     [label, score] = HCclassification_fit(weight, xbar, s, Test)

 Inputs:
   weight   p-by-1 vector that shows the weight for each predictor
   xbar     p-by-1 vector, estimated mean from HCclassification function
   s        p-by-1 vector, estimated standard deviation from HCclassification function
   Test     M-by-P matrix of predictors for data set to be predicted
   
 Outputs
   label    m-by-1 vector of estimated labels "1" or "0"
   score    m-by-1 vector, showing classification score for test data

 
 
LICENSE

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program.  If not, see <http://www.gnu.org/licenses/>.

If you use this code for your publication, please include a reference to the paper "Higher criticism thresholding: Optimal feature selection when useful features are rare and weak".
 
 
CONTACT

For any problem, please contact

Jiashun Jin
Department of Statistics 
Carnegie Mellon University 
Email: jiashun@stat.cmu.edu

Wanjie Wang 
Department of Statistics 
University of Pennsylvania 
Email: wanjiew@wharton.upenn.edu