recovering the delay

[Abolfazl Ziaeemehr]

Hi, 

I am following this article:

Momentary information transfer as a coupling measure of time series

and I could reproduce the results for Bidirectionally coupled logistic maps in using JIDT, but I stuck in the couple Lorenz oscillators and get something like this:

jarLocation = "./infodynamics.jar"

init_jvm(jarLocation)

method = "infodynamics.measures.continuous.kraskov"

calcClass = jp.JPackage(method).TransferEntropyCalculatorKraskov

calc = calcClass()

calc.setProperty("k", 4)

calc.setProperty("k_TAU", "1")

calc.setProperty("l_TAU", "1")

calc.setProperty("k_HISTORY", "1")

calc.setProperty("l_HISTORY", "1")

for i_delay in range(len(DELAYs)):

    calc.setProperty("DELAY", str(DELAYs[i_delay]))

    calc.initialise()

    calc.startAddObservations()

    for trial in range(num_trials):

        source = data["z1"][trial]

        target = data["z2"][trial]

        calc.addObservations(source, target)

    calc.finaliseAddObservations()

    result = calc.computeAverageLocalOfObservations()

The followed the parameters of the paper. integration step is 0.001, observation step is 50 (every 50 time step record the coordinates), I tried with noise and without noise and search for different delay values. 

delay: tau_12 = 1, which is equivalent to 50 time step. I search for delays 40 to 60. 

I use Z1 and Z2 for analysis of information measurements.

I don't know what I am doing wrong?

I  am using latest version of JIDT and my OS is ubuntu 18.04.

Thanks in  advance for any guide.

[Abolfazl Ziaeemehr]

I used the simulation data in TRENTOOL package that implemented this example,  here (https://github.com/trentool/TRENTOOL3_exampledata/tree/master/Lorenz_2_systems)

and I used jidt fot calculation of the TE.

the result with AUTO_EMBED_METHOD", "RAGWITZ" 

AUTO_EMBED_TAU_SEARCH_MAX="3",

AUTO_EMBED_K_SEARCH_MAX="3",

AUTO_EMBED_RAGWITZ_NUM_NNS="4"

the maximum value of TE was 0.055501 at 48.

The correct value of delay is 45.

and without autoembeding:

It that good enough?

I appreciate if someone know how to make the it better.