README

fvplay: Feature Vector Play

This set of MATLAB scripts is intended for visualizing and preparing raw data 
for use with an SVM classifier.  As is commonly known, the features that are used 
in the classifier are more important than the method itself, thus these tools 
provide an easy way to visualize and interpret data for best feature creation.  
Data is expected to be an n x m matrix of doubles with two classes.
Package includes:
			fvprep
			fvplay
			fvplot


%fvprep: Prepares data objects for use with rest of package. %%%%%%%%%%

USAGE: fvprep(rawdata,labels)
INPUT: rawdata: an n x m matrix of doubles, with n objects / subjects 
		each with m features
       labels: a single column vector of size n x 1 with limited to values of
	-1 and 1 to distinguish the two classes
OUTPUT: fv.features_vectorsraw data
	fv.index1 - indices in inputdata for label 1 (1)
	fv.index2 - indices in inputdata for label 2 (-1)
        fv.m      - all means of features (columns)
        fv.u1, fv.s1, fv.v1: svd output for group 1
	fv.u2, fv.s2, fv.v2: svd output for group 2


% fvplay takes as input an fv data object prepared by fvprep, and %%%
 allows the user to select a transform to plot (to explore the data)
 fvplot returns the data, and also displays plot in figures 1 and 2

 USAGE: fvplot(fvobject,type)
 INPUT: fvobject: data object prepared by fvprep
	 type: options include:
			quad: quadratic transformation
			exp: exponential transformation
			sig: sigmoid transformation
			log: logarithmic transformation
 OUTPUT: data: transformed n x m data matrix with n subjects (rows)
		and m features (columns)


% fvplot takes as input the output of fvplay (an n x m matrix with n %% objects / subjects (rows) and m features (columns) and plots a mean
 subtracted histogram and heat map.

 USAGE: fvplot(fvobject,type)
 INPUT: transform: transformed n x m data matrix, with n objects and m features
         fv: fv data object
 OUTPUT: graphical... plots!


% fvcorr takes as input a data matrix and associated labels (in format -1 and 1)
 and calculates the cross correlation matrix.
 USAGE: fvcorr(data, labels)
 INPUT: 
	 data: a n x M matrix with n subjects, m columns of features
    labels: a n X 1 column of labels, -1 and 1
 OUTPUT: ccdata: cross correlation matrix