TSTOOL home page | TSTOOL documentation page | TSTOOL link page

next up previous contents
Next: 6.9 fnearneigh - Fast Up: 6. Mex-Function Reference Previous: 6.7 corrsum - Computation

6.8 corrsum2 - Computation of the correlation sum

This is an extended version of the correlation sum algorithm. It tries to accelerate the computation of the correlation sum by using a different number of reference points at each length scale. For large length scales, only a few number of reference points will be used since for this scale, quite a lot of neighbors will fall within this range (and also the search time will be high). The smaller the length scale, the more reference points are used. The algorithm tries to keep the number of pairs found within each range roughly constant at Npairs to ensure a good statistic even for the smallest length scales. However, the number of reference points actually used may be limited to be within [Nref_min Nref_max] to give at least some control to the user. All reference points are chosen randomly from the data set without reoccurences of the same index.

Syntax:

Input arguments:

Output arguments:

Example:

x = chaosys(25000, 0.025,  [0.1 -0.1 0.02], 0);
x = x(5001:end,:);              % discard first 5000 samples due to transient
% now compute correlation sum up to five percent of attractor diameter
[c,d] = corrsum2(x,[1000 100 2000], 0.05, 200);
loglog(d,c)        % and show the result as log-log plot

next up previous contents
Next: 6.9 fnearneigh - Fast Up: 6. Mex-Function Reference Previous: 6.7 corrsum - Computation   Contents
TSTOOL

Copyright © 1997-2009 DPI Göttingen