gusucode.com > 支持向量机工具箱 - LIBSVM OSU_SVM LS_SVM源码程序 > 支持向量机工具箱 - LIBSVM OSU_SVM LS_SVM\stprtool\datasets\datafiles.m
% DATAFILES describes data files used in the STPR toolbox. % % The demo programs and other toolbox functions use two % types of data files: % 1) finite point sets. % 2) mixture of Gaussians. % % The data files can be created by the user program or % interactively by program 'creatset'. % % 1) Finite point sets: % ----------------------------- % are stored in Matlab file which must contain following variables: % % id [string] string identifier id = 'Finite sets, Enumeration'. % The identification string can be obtaind also by command % id=dataid(1). % X [NxL] matrix representing finite point set. The number % of points is L. The points are stored as N-dimensional % column vector, i.e. X = [x1,x2,...,xL]. % I [1xL] vector of integers which contains labels for points X. % The i-th point X(:,i) has label I(i). % Most functions use label 1 for the 1st class, % label 2 for the 2nd class and so on. % N [1x1] data dimension. % K [1xM] contains numbers of points in indicidual classes, % i.e. having the same label. The M is number of classes % (usualy M=max(I)). The integer K(i) is number % of points which belong to the i-th class, i.e. % the number of points which have label=i. % % Example: % id = dataid(1); % or id = 'Finite sets, Enumeration'. % X = [[0;0],[1;0],[0;1],[1;1]]; % logical AND. % I = [ 1 2 2 2 ]; % labels. % N = 2; % dimension is also size(X,1); % K = [1,3]; % also K(1)=length(find(I==1)), % % K(2)=length(find(I==2)) and % % sum(K) == size(X,2); % % 2) Mixture of Gaussians: % ---------------------------------- % is stored in Matlab file which must contain following variables: % % id [string] string identifier id = Infinite sets, Normal distributions'. % The identification string can be obtaind also by command % id=dataid(2). % MI [NxL] matrix which contains mean vectors. The number % of points is L. The mean vector are stored as N-dimensional % column vector, i.e. MI = [mi1,mi2,...,miL]. % SIGMA [Nx(L*N)] matrix which contains covariance matrices. The number % of matrices is L. The matrices are store one by one, i.e. % SIGMA =[sigma1,sigma2,...,sigmaL], where sigmai is i-th % covariance matrix. To take i-th matrix use command % acov(SIGMA,i). % I [1xL] vector of integers which contains labels. The i-th % Gaussian represented by the mean vector MI(:,i) and % covariance matrix acov(SIGMA,i) has label I(i). % N [1x1] data dimension. % K [1xM] contains numbers of Gaussians in individual classes, % i.e. having the same label. The M is number of classes. % The integer K(i) is number of Gaussians which have label i. % % Example: % id = dataid(1); % or id = 'Infinite sets, Normal distributions'. % MI = [[-1;-1],[1;1]]; % two Gaussians with mean vectors [-1;-1],[1;1] % SIGMA = [eye(2,2),eye(2,2)] % and identity covariance matrices. % I = [ 1 2]; % labels. % N = 2; % dimension is also size(MI,1) == size(SIGMA,1); % K = [1,1]; % also K(1)=length(find(I==1)), % % K(2)=length(find(I==2)) and % % sum(K) == size(MI,2) % % See also CHECKDAT, DATAID, CREATSET. % % Statistical Pattern Recognition Toolbox, Vojtech Franc, Vaclav Hlavac % (c) Czech Technical University Prague, http://cmp.felk.cvut.cz % Written Vojtech Franc (diploma thesis) 02.01.2000 % Modifications % 26-June-2001, V.Franc, created.