The Spider Objects

 
 
 
 
   
  Multi-class feature selection using spectral clustering
   
 


   A=CLUSTUB(H) returns a clustub object initialized with hyperparameters H. 
  
    Peforms feature selection via spectral clustering.
  
   Hyperparameters, and their defaults
    feat=[]              -- number of features
    output_rank=0        -- set to 1 if only the feature ranking matters
                            (does not perform any classification on the data)
    child=spectral('k=2')-- Clustering method used for training
  
   Model
    w                    -- the weights
    b0                   -- the threshold (when using all features)
    rank                 -- the ranking of the features
    d                    -- training set
  
   Methods:
    train, test, get_w 
  
   Example:
    d=gen(bayes({gauss([-1 3]) gauss([0 4]) gauss([1 2])})) 
    a=chain({clustub('output_rank=1'),one_vs_rest(svm('ridge=0.01'))})
    [r,a]=train(a,d)
  
   Note:
    Method for multi-class feature selection.
    To use with SVM, use:  
    chain({clustub('output_rank=1'),one_vs_rest(svm)})
  

the Emergence of Sparsity in a Weighted-based Approach.

Reference : Feature Selection for Unsupervised and Supervised Inference:

Author : Lior Wolf and A. Shashua

Link : http://www.cs.huji.ac.il/~shashua/papers/fts-long.pdf