read_sparse.m

libsvm-demo,支持向量机的演示程序,对初学者很有用!

%
% This reads a problem (in svmlight format) and
% return label vector and features in sparse matrix.
%

function [svm_lbl, svm_data] = read_sparse(fname); 

fid = fopen(fname); 
line = 0; 
elements = 0;
svm_lbl = []; 
svm_data = []; 

while 1 
	tline = fgetl(fid); 
	if ~ischar(tline), break, end 
	line = line + 1; 
	[lbl, data] = strtok(tline); 

	svm_lbl(line, 1) = sscanf(lbl, '%g'); 
	[c, v] = strread(data, '%d:%f'); 
	elements = elements + length(c);
end

frewind(fid);

row = zeros(elements,1);
col = zeros(elements,1);
value = zeros(elements,1);
elements = 1;
for i = 1:line
	tline = fgetl(fid); 
	if ~ischar(tline), break, end 
	[lbl, data] = strtok(tline); 

	svm_lbl(i, 1) = sscanf(lbl, '%g'); 
	[c, v] = strread(data, '%d:%f'); 
	s = elements;
	e = elements + length(c) - 1;	
	row(s:e, 1) = i * ones(length(c), 1);
	col(s:e, 1) = c;
	value(s:e, 1) = v;
	elements = elements + length(c);
end

svm_data = sparse(row, col, value); 

fclose(fid);