www.gusucode.com > C-C方法计算时间延迟和嵌入维数计算Lyapunov指数计算关联维数混沌时间序列预测 > code/Chaos Toolbox Ver.2.0/Main/largest_lyapunov_exponent.m

    function lambda_1=largest_lyapunov_exponent(data,N,m,tau,P)
%the function is used to calcultate largest lyapunov exponent with the
%mended algorithm,which put forward by lv jing hu.
%data:the time series
%N:the length of data
%m:enbedding dimention
%tau:time delay
%P:the mean period of the time series,calculated with FFT
%lambda_1:return the largest lyapunov exponent
%skyhawk
delt_t=1;
Y=reconstitution(data,N,m,tau );%reconstitute state space
M=N-(m-1)*tau;%M is the number of embedded points in m-dimensional space
for j=1:M
    d_max=1e+100;
    for jj=1:M                                              %寻找相空间中每个点的最近距离点，并记下
        d_s=0;                                              %该点下标
        if abs(j-jj)>P                                      %限制短暂分离
            for i=1:m
                d_s=d_s+(Y(i,j)-Y(i,jj))*(Y(i,j)-Y(i,jj));
                d_min=d_max;
                if d_s<d_min
                   d_min=d_s;
                   idx_j=jj;
               end
            end
        end
    end
%     index(j)=idx_j;
    max_i=min((M-j),(M-idx_j));%计算点j的最大演化时间步长i
    for k=1:max_i              %计算点j与其最近邻点在i个离散步后的距离
        d_j_i=0;
        for kk=1:m
            d_j_i=d_j_i+(Y(kk,j+k)-Y(kk,idx_j+k))*(Y(kk,j+k)-Y(kk,idx_j+k));
            d(k,j)=d_j_i;
        end
    end
end

%对每个演化时间步长i，求所有的j的lnd(i,j)平均
[l_i,l_j]=size(d);
for i=1:l_i
    q=0;
    y_s=0;
    for j=1:l_j
        if d(i,j)~=0
            q=q+1;
            y_s=y_s+log(d(i,j));
        end
    end
    y(i)=y_s/(q*delt_t);
end
x=1:length(y);
pp=polyfit(x,y,1);
lambda_1=pp(1);
yp=polyval(pp,x);
plot(x,y,'-o',x,yp,'--')