Working Definitions for the Design Detection Game/Tool

I want to thank OMagain in advance for doing the heavy lifting required to make my little tool/game sharable. His efforts will not only speed the process up immeasurably they will lend some much needed bipartisanship  to this endeavor as we move forward. When he is done I believe we can begin to attempt to use the game/tool to do some real testable science in the area of ID . I’m sure all will agree this will be quite an accomplishment.
Moving forward I would ask that in these discussions we take things slowly doing our best to leave out the usual culture warfare template and try to focus on what is actually being said rather than the motives and implications we think we see behind the words.

 

I believe now would be a good time for us to do some preliminary definitional housework. That way when OMagain finishes his work on the gizmo I can lay out some proposed Hypotheses and the real fun can hopefully start immediately.

 

It is always desirable to begin with good operational definitions that are agreeable to everyone and as precise as possible. With that in mind I would like to suggest the following short operational definitions for some terms that will invariably come up in the discussions that follow.

 

1.      Random– exhibiting no discernible pattern , alternatively a numeric string corresponding to the decimal expansion of an irrational number that is unknown to the observer who is evaluating it

2.       Computable function– a function with a finite procedure (an algorithm) telling how to compute the function.

3.       Artifact– a nonrandom object that is described by a representative string that can’t be explained by a computable function that does not reference the representative string

4.      Explanation –a model produced by a alternative method that an observer can’t distinguish from the string being evaluated

5.       Designer– a being capable of producing artifacts

6.       Observer– a being that with feedback can generally and reliably distinguish between artifacts and models that approximate them

Please take some time to review and let me know if these working definitions are acceptable and clear enough for you all. These are works in progress and I fully expect them to change as you give feedback.

Any suggestions for improvement will be welcomed and as always please forgive the spelling and grammar mistakes.

peace

541 thoughts on “Working Definitions for the Design Detection Game/Tool

  1. newton,

    This will sound snarky, but I seriously suspect the root of this issue is the authoritarian nature of the fundamentalist and evangelical churches. Questioning is dangerous and discouraged. That training affects believers’ behavior in secular environments

    I agree, it sounds snarky.

    But does it sound wrong?

  2. I think it’s wrong. I think fearfulness about ideas is a character trait rather than a property of theologies.

  3. petrushka,

    I think it’s wrong. I think fearfulness about ideas is a character trait rather than a property of theologies.

    Perhaps those with that trait are attracted to authoritarian religions that discourage questioning.

  4. petrushka:
    I’m still trying to get a thumbs up or down on how the “strings” are to be formatted.

    To make it easier to import something like a sonnet, I propose the following convention.

    For each unique word in the sonnet, assign a number.

    For example:

    Shall = 1
    I = 2
    compare = 3
    thee = 4
    to = 5
    a = 6
    summer’s = 7
    day = 8
    ?= 9

    so the first line of the sonnet, converted to an integer dataset, would be

    1,2,3,4,5,6,7,8,9

    Anyway, this convention could be used for any kind of object that has a reasonably small list of discrete items or values. Images, for example, or genomes.

    We could start the project without bickering about base conversion or offsets or whatever.

    I would say that all you have to do is convert the letters making up the sonnet into ASCII characters and string those together, comma separated. Make sure to include spaces and line breaks as ASCII too.

    The string will obey the rules of the English language and therefore not be random (in the sense that every character is equiprobable at a given position, because certain letters and letter combinations occur far more often than others). Up to a point there will be some predictability at local scale, again because of these rules.

    Fmm will then add random noise as per his workflow, and smooth it all out a bit again using his EA. The output is likely to be a bit more spiky than the original. I don’t know if the model will be different enough to raise the ‘design’ flag, but don’t worry, even if it doesn’t fmm will still consider the original designed because he believes that everything is designed!

    fG

  5. Ascii is fine by me. I asked a number of times how he wants the data presented.

    I guarantee that no matter how wonderful his analysis is, it will be possible to get false positives and negatives by generating data with an EA. He already knows this, because he linked to a paper showing that even sophisticated algorithms can be fooled.

    I have popcorn ready and am willing to play, but it looks like TimeCube stuff from here.

  6. Any limits or reasonable expectations for the size of the items or the length of test strings?

  7. keiths:

    As he’s described it, there is no score. It’s either “design inferred” or “design not inferred”:

    petrushka:

    That would be magic. There has to be a metric somewhere in the running of the code.

    No, because it is the observer, not the program, who decides whether to infer design. The code just presents line graphs of the strings to the observer.

    In fifth’s own words:

    In order to actually infer design the observer needs to be unable to distinguish the real string from the “manual” and “complexity” strings but able to distinguish when it comes to the “model” and “random” strings.

  8. petrushka:
    Any limits or reasonable expectations for the size of the items or the length of test strings?

    There seem to be 40 lines that can be drawn at a time, lines 98 to 178 in the code.

  9. petrushka:

    I guarantee that no matter how wonderful his analysis is, it will be possible to get false positives and negatives by generating data with an EA. He already knows this, because he linked to a paper showing that even sophisticated algorithms can be fooled.

    The issue is that fmm considers the output of EA’s designed. He has said so to me upthread. The way he frames the problem there will never be any false positives because according to him everything is designed.

    fG

  10. I’ve downloaded the processing app, but need a bit of help figuring out how to load and run FMM’s app.

  11. faded_Glory: I ran your string through a simple time series analysis program (caveat: I am no professional statistician) and this is what I found:

    Excellent, That is exactly the sort of honest effort that I’m beginning to appreciate with you. Thanks

    Your appraisal is mostly spot on

    faded_Glory: Natural data comes in all sorts, smooth, noisy, whatever, depending on what it is, how it originates and how you measure it. I see no reason to label smooth datasets as ‘designed’ and more noisy ones as ‘undetermined’ unless you throw in an awful lot of other assumptions.

    We are not looking for smooth data sets. It just so happens that this how this particular data set presents itself.

    Real strings will differ from models in different ways sometimes they will be smother sometimes more choppy sometimes they will contain no plateau and other times there will be more plateaus that you would expect given the data.

    Every data set is different. That is the reason that you can’t design a general purpose algroythym to model these strings.

    You can get close with an EA but close does not cut the mustard in this regard. With a little feedback you can always distinguish between the real and the model

    recall Maguire’s definition of an integrating function

    quote:
    the knowledge of m(z)does not help to describe m(z′),when z and z′are close
    End quote:

    I would argue that is what we see with these strings

    Early on I would try to tweak the EA to get a better match but I quickly found that the algroythym just got more and more clunky and differences remained regardless and often got more pronounced.

    Thanks again for the interaction

    peace

  12. OMagain,

    Do you know how I can output the required format from Excel? I have a file I want to use for testing, it is a series of numbers with each number in a separate cell (ordered as a row or as a column). Somewhere between 300 and 400 numbers.

    I tried space delimited (*.prn) output and renamed that to real.txt and fake.txt ( the same file for now) but the program can’t read it.

    Thanks,

    fG

  13. faded_Glory: By the way, when I construct an correlogram of your original data with 106 lags there appears a periodicity at lags of 30 and multiples of that. Is there a seasonal fluctuation, or something like it, in your data?

    Again excellent observation,

    I hesitate to confirm or deny because part of the fun in looking at these stings is thinking about how much you can learn from examining and comparing raw data in the complete absence of context.

    It’s almost like you are learning the global pattern via lossless information integration. 😉

    You could test your hypothesis with the game by choosing the string with lags at multiples of thirty and see if you are correct

    If I spill the beans to quickly it means we will have to start over with a new string

    peace

  14. faded_Glory: Do you know how I can output the required format from Excel?

    I’m not in the office now but I could send you the original Excel sheet that I cobbled together when I make it back.

    IMO it is better than the program.

    I’m confident that in the end OMagain will put something together that is elegant and practical but as it stands the program needs lots of work.

    It is after all my first attempt at programming

    peace

  15. I’m trying to use the sample files and get:

    processing.app.SketchException: unexpected char: ‘D’
    at processing.mode.java.JavaBuild.preprocess(JavaBuild.java:399)
    at processing.mode.java.JavaBuild.preprocess(JavaBuild.java:193)
    at processing.mode.java.JavaBuild.build(JavaBuild.java:152)
    at processing.mode.java.JavaBuild.build(JavaBuild.java:131)
    at processing.mode.java.JavaMode.handleLaunch(JavaMode.java:153)
    at processing.mode.java.JavaEditor$36.run(JavaEditor.java:1099)
    at java.lang.Thread.run(Thread.java:745)

    While opening fake.txt

  16. keiths: No, because it is the observer, not the program, who decides whether to infer design. The code just presents line graphs of the strings to the observer.

    Right there is nothing special about the program it just provides a platform to compare. I’ve even toyed with other ways to present the data that don’t involve line graphs.

    We could for example try pulsating balls that vary in size according to the values of each number in the sequence.

    All that is important is that we be able to compare the strings visually.

    peace

  17. fifthmonarchyman:

    Real strings will differ from models in different ways sometimes they will be smother sometimes more choppy sometimes they will contain no plateau and other timesthere will be more plateaus that you would expect given the data.

    Every data set is different. That is the reason that you can’t design a general purpose algroythym to model these strings.

    You can get close with an EA but close does not cut the mustard in this regard. With a little feedback you can always distinguish between the real and the model

    I would argue that is what we see with these strings

    Early on I would try to tweak the EA to get a better match but I quickly found that the algroythym just got more and more clunky and differences remained regardless and often got more pronounced.

    What I think is going on is that your randomisation of the original string introduces a lot of noise that you can never fully remove anymore. I would have to dig deep to find the correct explanation, but there is an upper limit to S/N in digital signals (and that is what your strings are).

    Therefore, your model will always be noisier than your original, and if you zoom in closely enough you will be able to see that. Depending on the smoothness of your original data this will be easier or harder to do.

    It is a bit like digital audio – digital music will never be identical to the original analog sound. The reason it sounds good is that the sampling frequency has been set so high that the noise is pushed into that part of the spectrum where our ears can’t hear it anymore. At the same time other sources of signal degradation have been eliminated by the recording process so that the music sounds very ‘clean’ compared to analog recordings. But it will never be exactly the same, and if you had unlimited resolution powers in your hearing you would be able to tell the difference.

    In your game you can set the viewing resolution so high that you can view every single sample, and therefore you will always be able to spot the difference between the original and the noisy copy.

    I honestly think that this is all that is going on here, and all the rest about ‘integrated information’ is just a red herring.

    fG

  18. faded_Glory: Therefore, your model will always be noisier than your original, and if you zoom in closely enough you will be able to see that. Depending on the smoothness of your original data this will be easier or harder to do.

    Actually I have several real strings that fluctuate widely from point to point more so than you would expect by chance

    In those strings the model is smother than the original

    faded_Glory: In your game you can set the viewing resolution so high that you can view every single sample, and therefore you will always be able to spot the difference between the original and the noisy copy.

    The graphs go by rather quickly and you are severely limited in regards to time.

    You have to choose quickly

    This prevents you from viewing every single sample you have to go with the overall pattern, that is the theory anyway

    peace

  19. petrushka:
    I’m trying to use the sample files and get:

    processing.app.SketchException: unexpected char: ‘D’at processing.mode.java.JavaBuild.preprocess(JavaBuild.java:399)at processing.mode.java.JavaBuild.preprocess(JavaBuild.java:193)at processing.mode.java.JavaBuild.build(JavaBuild.java:152)at processing.mode.java.JavaBuild.build(JavaBuild.java:131)at processing.mode.java.JavaMode.handleLaunch(JavaMode.java:153)at processing.mode.java.JavaEditor$36.run(JavaEditor.java:1099)at java.lang.Thread.run(Thread.java:745)

    While opening fake.txt

    A shot in the dark, but it could be the BOM in UTF-8?

  20. try just a file with

    1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9

    on a single line

  21. OMagain:
    try just a file with
    1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
    on a single line

    I tried that.

    I also cannot figure out where it is looking for the txt file. I tried setting preferences, but it doesn’t find it unless I hard code the path in the program.

  22. petrushka: I also cannot figure out where it is looking for the txt file. I tried setting preferences, but it doesn’t find it unless I hard code the path in the program.

    Put it in the same place as the processing executable. It then just magically finds it.

  23. petrushka: Or maybe the POS in the java.exe file.

    LOL
    Well, it’s working here, tho I have no clue what to do with it
    I tried with a fake.txt encoded in UTF-8 with BOM and it did crap out here too, but I got a different exception (NumberFormat)

  24. Okay, the unexpected character is in the filename. Where the fuck is it looking for the txt file? It isn’t looking in the folder specified in preferences.

  25. dazz: LOL
    Well, it’s working here, tho I have no clue what to do with it
    I tried with a fake.txt encoded in UTF-8 with BOM and it did crap out here too, but I got a different exception (NumberFormat)

    The line moves across, you “guess” which one is the fake and the real string by clicking on the respective buttons.

    I think 😛

  26. Okay, the mechanics are done. Now it appears that the “real” string is always on top. Is this just me, or is this a beta version?

  27. fifthmonarchyman: Actually I have several real strings that fluctuate widely from point to point more so than you would expect by chance

    In those strings the model is smother than the original

    First of all, saying something is ‘fluctuating more wildly than you would expect by chance’ sounds very sloppy. How do you determine what the ‘expected fluctuation’ is of a ‘chance string’?

    I don’t know how you randomise your original, but I would think that this step will in all cases add noise to the original unless that was already pretty much white noise to begin with. Likewise, I don’t know what your GA does exactly, but I find it hard to understand how it would result in a smoother fake if you start with a randomised copy of a spiky (i.e. noisy) string.

    We need to see more examples, and more detailed explanation of the steps you take to produce your fakes. OMagain’s program doesn’t do any of that, right? It just does the display thing. Tell us exactly what you do to your data.

    The graphs go by rather quickly and you are severely limited in regards to time.

    You have to choose quickly

    This prevents you from viewing every single sample you have to go with the overall pattern, that is the theory anyway

    More observer-dependent influences on the result. You are still very far from science.

    fG

  28. OMagain: The line moves across, you “guess” which one is the fake and the real string by clicking on the respective buttons.

    I think

    What if I swap the files? this makes no sense to me
    If “fake” is simply determined by what’s in the fake.txt, and one can swap the contents of fake & real.txt, it turns the game into a 50-50 thing, very much like “guess which hand”

  29. I’ll start with a first effort. Here are two strings, one of which is an actual dataset from published science , and the second is a randomized dataset of the same size.

    40 10 10 20 20 10 40 20 30 10 10 10 10 30 20 10 10 20 30 40 10 20 20 40 40 20 10 10 10 20 40 40 10 40 20 20 30 30 40 20 10 20 40 10 10 20 40 40 20 30 20 40 30 20 40 40 30 20 30 10
    20 40 30 40 40 10 30 40 20 30 20 20 20 20 10 30 20 30 10 10 10 10 10 40 40 40 20 20 20 10 20 20 40 30 20 40 10 10 10 40 40 10 30 20 20 10 10 20 40 30 20 30 10 40 20 40 10 30 20 20
    20 20 30 10 20 40 10 20 10 40 30 10 20 30 20 20 30 30 10 20 10 20 20 10 20 10 40 40 40 40 30 40 40 40 10 10 20 10 40 10 40 10 10 40 20 40 10 20 40 10 20 40 20 20 20 20 40 40 40 10
    10 40 10 40 40 10 20 20 20 20 10 10 20 40 30 20 30 10 40 40 40 30 20 20 30 20 20 30 10 10 20 20 10 40 30 10 20 10 20 20 30 30 10 20 10 20 20 10 20 10 40 40 40 40 40 40 20 40 10 40
    10 10 40 10 10 40 10 10 10 40 10 20 40 10 30 10 30 30 20 10 40 20 40 30 20 20 20 10 20 10 20 10 40 40 10 20 10 20 10 30 40 40 30 20 40 30 20 30 40 30 40 30 10 40 10 20 40 10 30 40

    40 40 10 40 20 10 30 20 10 20 10 40 10 40 30 10 40 40 40 10 20 40 20 10 20 40 10 40 20 20 10 10 10 30 40 10 10 40 30 30 30 10 30 30 30 10 40 10 10 30 40 30 40 40 40 10 40 20 20 30
    10 30 20 10 20 40 10 40 40 10 10 30 10 10 20 10 40 40 30 20 20 10 40 20 10 30 10 40 30 20 20 30 30 20 40 20 10 20 40 30 20 40 30 10 10 10 10 20 40 20 30 10 30 20 20 30 40 40 10 40
    20 10 20 30 10 10 10 30 40 30 20 10 20 20 20 20 20 40 30 10 20 10 10 10 40 20 10 10 40 10 30 40 20 40 40 30 30 40 30 20 40 20 10 40 10 20 10 20 30 20 40 10 30 10 20 10 40 20 30 20
    40 40 40 30 30 30 30 10 30 10 10 40 40 10 10 20 10 40 40 40 40 30 20 10 20 40 40 20 40 40 40 40 40 20 30 20 10 20 30 40 20 30 10 20 20 40 20 20 20 30 30 20 20 10 30 10 10 40 20 10
    10 20 40 30 30 10 30 40 10 40 40 40 20 40 10 20 30 30 20 30 40 10 10 10 40 40 30 20 10 20 30 10 20 10 10 40 10 10 30 10 30 10 30 30 30 10 10 20 30 30 40 40 10 40 40 10 30 20 10 10

  30. My turn!

    One of these two is an amino acid sequence of a human protein found at uniprot.org, each amino acid translated to a number from 1 to 22
    The other one is a randomly generated sequence with Python with numbers ranging 1 to 22:

    12 7 5 3 18 2 7 11 11 3 21 7 19 12 16 3 22 11 11 17 2 20 17 16 6 16 16 4 8 12 8 8 17 19 20 6 20 22 15 22 3 9 20 9 16 10 16 4 14 6 3 10 1 3 20 1 15 6 21 15 6 10 16 1 3 22 20 3 18 21 1 18 8 11 13 3 11 21 6 1 9 22 9 12 16 13 1 1 5 14 21 5 8 9 13 6 3 4 20 11 19 22 10 7 8 15 15 19 11 15 2 20 14 3 7 22 15 5 19 6 10 12 12 16 4 13 12 1 2 22 12 17 7 12 21 12 16 10 14 16 2 16 6 18 15 5 4 2 20 5 10 21 6 22 20 18 1 4 12 12 9 3 14 19 15 19 16 1 10 9 15 17 2 17 9 1 21 3 17 14 4 1 19 13 19 22 14 4 2 5 22 14 22 3 8 17 11 20 21 1 18 13 4 19 21 4 3 1 9 12 4 2 10 10 22 7 17 5 4 5 7 10 1 5 18 6 1 5 16 22 19 16 13 12 19 20 19 15 22 6 19 2 11 18 15 12 3 15 9 22 18 15 18 12 16 7 5 18 8 21 17 19 9 21 9 15 9 4 4 15 13 9 9 11 4 16 7 20 6 10 3 20 11 12 20 16 12 8 11 13 20 13 3 10 12 22 15 2 11 15 22 22 2 18 9 1 11 7 7 2 4 18 21 15 15 9 17 15 9 18 8 8 9 5 9 14 17 18 9 16 12 20 4 10 19 13 14 12 2 13 21 9 12 1 17 17 21 11 5 6 4 10 16 22 7 11 11 18 4 10 18 8 4 13 11 15 18 16 1 7 20 18 8 2 8 22 7 8 10 7 16 19 4 2 22 6 3 16 17 2 20 8 16 22 5 13 13 21 15 14 19 14 22 11 3 9 22 7 3 4 12 1 3 17 6 21 8 11 10 16 7 1 11 7 2 17 19 1 14 2 15 10 2 4 5 15 3 10 16 20 5 6 9 3 3 19 8 4 21 12 20 2 17 9 2 15 7 22 19 18 4 11 17 21 7 13 21 9 11 19 11 1 4 2 18 20 16 8 2 22 9 3 18 11 7 14 4 10 5 18 6 14 10 14 22 9 18 19 1 1 3 5 13 12 17 2 17 22 19 9 8 1 18 8 2 6 8 8 14 12 15 22 7 8 8 16 9 2 2 15 13 17 5 8 19 16 12 7 7 14 6 18 3 12 16 16 4 9 1 2 5 15 9 13 3 11 10 21 6 7 18 4 12 7 2 11 13 11 20 11 19 21 18 1 7 11 13 7 2 17 4 11 20 8 1 21 6 1 19 1 18 22 11 14 20 16 8 2 13 9 10 1 6 14 14 4 6 16 7 8 10 16 10 10 21 12 1 8 2 9 13 4 17 21 19 5 4 15 17 4 11 7 9 14 3 16 12 15 7 16 10 4 9 3 12 8 8

    15 7 11 12 8 10 1 20 14 19 12 18 3 10 16 10 16 14 4 1 22 16 4 10 18 18 6 12 18 17 19 12 22 8 8 16 10 21 8 2 2 1 18 4 4 10 14 13 19 4 17 18 14 19 18 3 19 12 2 22 16 13 17 3 14 8 2 19 22 22 3 22 2 3 10 15 18 13 11 4 6 13 15 14 1 13 14 22 2 10 13 8 17 7 6 6 1 22 16 2 13 13 11 7 11 14 10 14 14 1 2 13 4 20 3 19 4 1 1 18 13 12 10 7 7 13 8 22 4 16 13 4 11 22 17 13 19 19 11 3 16 1 2 14 19 16 13 14 13 1 16 6 4 12 6 8 14 16 13 13 3 10 16 2 6 8 19 6 10 21 14 16 11 7 11 6 18 19 14 22 17 19 15 6 22 4 20 18 3 12 2 8 13 13 13 16 17 3 18 19 12 10 4 18 10 22 17 1 13 17 18 13 19 15 2 2 15 2 7 18 22 18 2 15 17 22 18 18 8 11 2 21 18 19 17 11 1 16 19 16 3 19 18 18 17 18 18 7 10 18 13 18 8 2 8 2 18 19 18 19 17 3 22 11 15 22 18 19 19 13 17 22 4 18 2 15 12 7 4 1 12 2 18 11 18 7 18 1 18 17 18 1 13 18 18 18 17 3 3 13 18 17 19 10 20 18 8 17 14 19 17 22 17 1 8 2 7 2 1 17 22 18 10 19 8 7 14 3 14 12 2 17 2 10 8 2 4 18 18 21 21 20 7 12 7 1 18 7 22 15 13 18 19 2 12 10 18 10 18 16 10 19 22 21 14 10 14 20 11 10 4 22 1 22 14 12 13 14 22 22 4 17 19 17 7 8 16 8 1 16 2 3 7 22 1 22 13 2 14 19 2 11 22 3 12 13 13 16 15 10 21 15 19 14 4 3 13 1 12 22 19 8 20 6 7 10 18 18 13 21 14 11 13 11 22 8 7 19 14 16 8 15 16 8 13 12 4 12 1 2 8 19 1 8 10 15 4 21 13 11 1 14 3 12 12 11 2 4 15 14 18 3 3 12 16 13 11 7 10 13 19 22 14 12 10 4 16 10 13 1 19 22 14 18 2 20 18 10 18 8 8 22 7 8 17 19 10 18 22 13 20 15 1 17 7 22 12 2 15 8 4 3 3 17 16 18 16 8 18 4 22 21 18 21 10 12 22 13 21 7 13 15 19 10 7 13 17 21 18 11 12 3 3 2 4 8 12 12 16 15 22 10 2 10 21 1 18 17 4 13 18 14 13 21 14 3 6 17 14 1 15 14 2 13 22 1 4 6 22 14 14 22 14 7 7 2 17 13 16 17 8 12 13 18 18 12 7 13 13 8 11 18 13 17 14 12 3 2 18 1 18 7 17 18 13 11 2 1 1 11 19 7 4 12 3 1 6 19 13 19 19 18 17 2 13 17 22 16

  31. fifthmonarchyman,

    I have focused exclusively on the contents of the paper. Either point out where the paper supports your vague allusions or admit that you are not talking about the same process.

    That is the point the paper is simply a summary of a game, You need to play the game to understand the paper. The authors of the paper have said as much and made the game available to anyone.

    Untrue. While the paper overstates its conclusions, it does actually describe the process followed in enough detail to replicate it. You could learn a lot from that writing style.

    As I summarized earlier, adapting that process to use machine learning rather than human participants would look like this:

    1) Download all historical closing prices for the DJIA.

    2) Randomly, with a uniform distribution, select a few thousand start dates.

    3) From each start date load the next 250 closing prices and save these time series.

    4) For each time series generate a permuted version following the algorithm described in the paper.

    5) Divide the set of time series pairs into training, cross validation, and test sets.

    6) Train a machine learning model using the training and cross validation sets by presenting pairs of real and permuted time series.

    7) Test the trained model on the test set.

    You have yet to explain why following this process, exactly that described in the paper, and achieving a 73% or better accuracy, exactly what humans achieved, would not meet your criteria for disproving your claims.

    Why don’t you play it and see what you are up against?

    Your game is not the same as that described in the paper and you have not yet provided sufficient clarity on the rules for anyone to be able to play it.

    I’ve described in detail what I propose. Either commit to abiding by the results or explain why following the exact process described in the paper doesn’t meet your criteria.

Leave a Reply