next up previous contents
Next: MLDK2MCC Up: Implementation Previous: MLDK2MCE   Contents


This program is similar to MLDK2MCE, but uses distances obtained from independent time series generated for every embedding dimension and every Monte Carlo trial. A time series is generated, stored in memory and checked for (exact) periodicity. There are some values for $x_0$ (e.g. $x_0 = 0$ for the logistic map) for which the solution becomes constant. In such a case, the initial condition is not attracted to the one we are interested in! Therefore, the time series will be discarded when the last iterate equals any of the previous ones. The values of the time series are transformed so that the maximum possible distance will equal 1 (so the reference distance will depend on the realization thereby introducing small additional fluctuations in the entropy estimates); distances are calculated using the supremum norm. The indices for the vectors (2.6) are randomly chosen with or without replacing. For the former case, see MLDK2; for the latter case, first all indices are given a random permutation (scrambled).