PARAFAC源程序，可以用于平行因子分析处理的算法，很全很好用(unction [A,B,C,LLF,I,J,K] = parafac(XPK,I,N,epsilon)
% PARAFAC Parallel factor analysis for an three-way data array
% The iterative process is continued until that m > 300 or ABS((LF(m)-LF(m-1))
% /LF(m-1)) is less than epsilon.Noted that m is the iterative number.
% The epsilon may also be 10*MAX(SIZE(XPK))*NORM(XPK,1)*EPS.
%
% This can be overridden with [A,B,C] = PARAFAC(XPK,I,N,epsilon).
%
% [A,B,C,LLF] = PARAFAC(XPK,I,N,epsilon) also returns a vector LLF(=LOG(LF))
% where LF is a loss function vector obtained by calculating the square of
% Frobenius matrix norm of (XPK-XPK*) at each iteration step.)