FMM design tool post 1

This is a temporary post so I can discuss the design tool with fmm, as trying to have a conversation across multiple threads is impossible.

45 thoughts on “FMM design tool post 1

  1. fifthmonarchyman: The whole point of making the Tool shareable is so that we can compare strings together. We need a way to input strings ourselves

    While that is true, a shared starting point is desirable. So when the tool starts, it has a demo ready to go. This is also desirable for other reasons. I.E. we start the tool and it behaves the same way for everybody initially.

    So, please generate the initial dataset and post it in a couple of comments on this thread.

    fifthmonarchyman: next put the randomized string through an EA till it is “close” to the original

    Another reason for wanting a starting dataset generated by you is this does not make much sense to me.

    How “close” is close? Why is an EA needed? What does it do, specifically? Does your existing code do that work? Can’t you just move the string randomly for a bit? How much is too much?

    Is the user supposed to know which strings are what before the test starts?

  2. I don’t have access to the excel version of the tool today. I’ll see if I can grab something when I get back to the office next week.

    peace

  3. OMagain,

    Why is an EA needed? What does it do, specifically?

    I asked a similar question lo these many moons ago:

    1) create a copy of the randomized string with a small number of random mutations
    2) compare both strings with the original
    3) discard the string with the smallest R squared
    4) repeat until a string reaches 80%

    How is this different from just randomizing bits in the string until the Hamming distance is equal to 0.2 * length?

    And again, what exactly does it model?

    fifthmonarchyman responded:

    I’m not a programer but a quick Google check seems to confirm that the concepts are equivalent for our proposes . I would need to study to be sure.

    You say,

    And again, what exactly does it model?

    If by “it” you mean the evolved copy it models the original string

    So the EA is an unnecessary flourish and I don’t think fifthmonarchyman knows what the whole process is supposed to model.

  4. Patrick: Why is an EA needed? What does it do, specifically?

    We are trying to create a string that is close to the “real” one but that is produced by an algorithmic process an EA is a great way to do this.

    Patrick thinks that another approach would be simply to take the original string and vary it randomly by a certain percent.

    I realize that (randomly very the original string by 20%) is an algorithm but it smells a little to arbitrarily “random” for my tastes.

    The benefits of an EA is that we start with complete randomness and approach the target of original string as defined by the r squared.

    We can stop at a predetermined value or continue to run to whatever r squared value we want.

    It’s just a better way of judging closeness in my view

    peace

  5. Without any selection or generations, what do you think the difference between the two methods is, FMM?

  6. fifthmonarchyman,

    We are trying to create a string that is close to the “real” one but that is produced by an algorithmic process an EA is a great way to do this.

    Why? You’re just stopping when the new string is 20% different from the old string. It doesn’t matter how you get there.

  7. Also keep in mind that I use this tool mainly to evaluate forecasting models.
    I run my model and then see if I can distinguish the forecasted results from the actual data using the tool.

    The metric that I generally use to evaluate regression equations is r squared. So I have two metrics that I can compare and contrast the binary one I get from the tool and the more established statistical one.

    What I find interesting is that at times I have a very good model in terms of the r squared and at the same time I can easily distinguish it from the original data.

    That is power of the tool

    peace

  8. Patrick: Why? You’re just stopping when the new string is 20% different from the old string. It doesn’t matter how you get there.

    You need to define what you mean by 20% different.

    IMO Randomly varying 20% of the digits is not the same as varying all the digits until l I get an r squared of .8.

    Another cool thing about this game is that we could try it both ways run a blind test and compare the results we get.

    Ah science

    peace

  9. fifthmonarchyman,

    Here’s what you said was your approach:

    1) create a copy of the randomized string with a small number of random mutations
    2) compare both strings with the original
    3) discard the string with the smallest R squared
    4) repeat until a string reaches 80%

    Regardless if your metric is 20% difference or r square of 0.8, I don’t see how your use of an EA is necessary. Does it contribute anything that simply randomly flipping bits would not?

  10. I think I sort of agree with Patrick but I think that flipping bits on the final string is missing the point. Generate a string at random and flip bits on it, including adding and subtracting characters. Or use an EA.

  11. Mung: Generate a string at random and flip bits on it, including adding and subtracting characters.

    I could agree with that

    peace

  12. OMagain: Is the user supposed to know which strings are what before the test starts?

    no.

    You need to choose which string is the “real” one by looking at the moving graph . Then check your answer.

    the buttons are supposed to turn green if you pick correctly

    After a few repetitions you should be able to do this with ease and be correct every time. At that point you know the “form” of the string.

    Hope that helps. Feel free to make any modifications you would like to make it all more elegant and user friendly

    I do thank you for your work on this.

    Coding is simply outside my wheelhouse and about as much fun as a root canal

    peace

  13. I love coding. I love it even more when given the time to do it the right way. But that rarely happens.

    Red -> Green -> Refactor

  14. Here is an original string that we can use as a base line. It is data recorded from a industrial type process

    51
    55
    57
    63
    66
    69
    72
    73
    75
    75
    76
    74
    74
    80
    77
    78
    76
    76
    75
    75
    75
    80
    80
    80
    82
    83
    83
    80
    83
    84
    82
    83
    84
    83
    84
    83
    83
    81
    81
    82
    82
    78
    78
    77
    80
    79
    76
    77
    78
    78
    77
    79
    80
    81
    83
    83
    72
    73
    73
    72
    70
    69
    69
    67
    64
    67
    69
    71
    72
    77
    78
    79
    79
    80
    81
    83
    84
    84
    85
    85
    85
    85
    84
    85
    84
    86
    87
    87
    87
    85
    85
    85
    84
    82
    82
    83
    84
    83
    83
    84
    85
    85
    84
    83
    81
    82
    82
    84
    86
    85
    85
    86
    86
    84
    84
    84
    84
    83
    82
    83
    83
    82
    79
    79
    80
    80
    79
    76
    77
    77
    78
    78
    79
    79
    80
    80
    82
    83
    83
    83
    83
    82
    84
    84
    85
    85
    84
    85
    84
    85
    85
    85
    85
    85
    86
    83
    83
    83
    83
    84
    84
    83
    83
    83
    83
    83
    81
    78
    69
    70
    71
    72
    74
    74
    70
    70
    71
    69
    70
    71
    71
    72
    74
    72
    70
    69
    70
    71
    71
    67
    66
    69
    69
    70
    70
    71
    72
    73
    74
    74
    73
    75
    76
    79
    80
    80
    82
    81
    80
    82
    82
    84
    84

  15. Here is the same data set but randomized as to order

    79
    70
    87
    79
    78
    79
    72
    72
    69
    78
    82
    69
    78
    86
    74
    83
    82
    83
    82
    83
    85
    85
    69
    83
    84
    72
    83
    80
    77
    82
    71
    83
    75
    70
    71
    81
    83
    81
    78
    86
    83
    83
    85
    86
    79
    77
    69
    85
    79
    81
    78
    85
    86
    75
    72
    77
    74
    82
    84
    84
    85
    81
    83
    84
    83
    83
    70
    74
    85
    83
    84
    71
    77
    73
    82
    83
    71
    66
    73
    83
    87
    80
    76
    79
    73
    72
    80
    78
    51
    83
    82
    85
    83
    82
    75
    80
    82
    80
    79
    74
    70
    85
    57
    80
    73
    81
    81
    82
    78
    78
    79
    64
    84
    80
    85
    66
    84
    83
    76
    69
    83
    83
    80
    83
    77
    82
    67
    81
    81
    77
    85
    83
    85
    71
    83
    83
    85
    84
    82
    76
    85
    80
    76
    82
    83
    77
    71
    69
    79
    74
    80
    83
    83
    82
    84
    83
    83
    77
    80
    84
    78
    78
    55
    84
    81
    78
    82
    85
    81
    84
    81
    74
    70
    80
    85
    80
    63
    67
    80
    70
    75
    82
    84
    83
    83
    80
    82
    72
    85
    71
    81
    82
    83
    84
    80
    78
    77
    83
    84
    83
    85
    87
    76
    82
    76
    85
    80
    83
    70
    83
    81
    70
    84

  16. To obtain this string I started with the random order set and ran it through a crude EA untill the R squared was over .85

    28
    37
    42
    53
    64
    58
    64
    70
    74
    67
    80
    67
    89
    72
    77
    84
    75
    73
    72
    69
    85
    74
    76
    85
    73
    77
    75
    92
    80
    83
    95
    74
    89
    79
    84
    85
    76
    84
    84
    90
    75
    77
    68
    90
    77
    66
    85
    71
    79
    68
    82
    87
    82
    89
    82
    50
    71
    63
    70
    65
    62
    58
    40
    44
    55
    46
    67
    63
    79
    77
    81
    77
    75
    80
    88
    79
    89
    90
    91
    88
    92
    96
    91
    91
    87
    96
    91
    99
    97
    94
    93
    86
    91
    87
    80
    93
    92
    88
    84
    91
    92
    81
    88
    78
    91
    86
    94
    90
    81
    99
    89
    94
    90
    83
    92
    90
    87
    81
    92
    90
    82
    76
    84
    69
    80
    75
    75
    77
    81
    79
    80
    81
    89
    80
    81
    89
    80
    91
    81
    81
    89
    94
    94
    95
    88
    89
    93
    86
    90
    97
    95
    95
    98
    96
    83
    91
    89
    76
    88
    95
    84
    84
    91
    92
    89
    88
    80
    68
    63
    55
    75
    64
    62
    58
    65
    53
    53
    73
    55
    59
    69
    64
    64
    72
    51
    51
    62
    67
    58
    49
    58
    53
    66
    66
    63
    61
    59
    61
    77
    69
    76
    70
    74
    77
    79
    86
    79
    78
    82
    93
    76
    96
    90

  17. Finally at Patrick’s suggestion I began with the original string and randomly varied 20% of the digits with a number that fell between the min and max of the set as a whole

    51
    68
    57
    63
    66
    56
    60
    73
    54
    75
    71
    74
    74
    62
    77
    78
    76
    76
    75
    75
    75
    80
    57
    80
    82
    83
    83
    80
    83
    84
    82
    83
    81
    83
    84
    63
    83
    81
    81
    82
    82
    78
    78
    77
    73
    79
    76
    77
    65
    80
    77
    79
    80
    81
    83
    83
    72
    73
    73
    72
    70
    69
    69
    67
    78
    67
    69
    71
    72
    77
    78
    79
    79
    80
    81
    58
    84
    84
    85
    85
    78
    85
    84
    85
    72
    86
    87
    87
    87
    85
    85
    85
    84
    82
    82
    83
    84
    83
    83
    79
    85
    85
    84
    83
    81
    82
    82
    74
    76
    85
    85
    86
    86
    84
    84
    84
    84
    83
    85
    83
    83
    82
    60
    79
    77
    80
    79
    76
    77
    81
    78
    67
    79
    79
    80
    80
    82
    64
    83
    83
    67
    82
    84
    84
    85
    85
    84
    85
    84
    85
    85
    85
    85
    85
    86
    83
    83
    83
    71
    84
    84
    83
    83
    83
    83
    83
    81
    78
    76
    70
    71
    72
    74
    74
    70
    70
    61
    69
    70
    71
    71
    63
    74
    58
    70
    69
    70
    71
    71
    67
    66
    62
    69
    70
    70
    81
    72
    73
    74
    74
    73
    75
    76
    79
    80
    54
    82
    81
    80
    76
    82
    84
    84

  18. Patrick: Could you please use Pastebin or something similar (or at least put your data comma separated on one line) rather than spamming the board?

    Sorry about that I’m done now

    peace

  19. some food for thought while we wait for Omagain

    http://www.nature.com/news/paradox-at-the-heart-of-mathematics-makes-physics-problem-unanswerable-1.18983

    quote:

    But the undecidability ‘at infinity’ means that even if the spectral gap is known for a certain finite-size lattice, it could change abruptly — from gapless to gapped or vice versa — when the size increases, even by just a single extra atom. And because it is “provably impossible” to predict when — or if — it will do so, Cubitt says, it will be difficult to draw general conclusions from experiments or simulations.

    end quote:

    peace

  20. Neil Rickert: If that is so, then quantum physics has gone badly wrong.

    If I was Patrick I’d demand you put together OP with a detailed argument defending your claim or that you retract it immediately in face of Hitchens’ mighty razor.

    But since I’m FMM I’ll just ask you how you know this

    😉

    peace

  21. fifthmonarchyman: But since I’m FMM I’ll just ask you how you know this

    Gödel’s incompleteness theorem is a very technical result about mathematical logic. It has no implications for finite logic, nor for empirical data. If it affects quantum physics, then quantum physics is not paying sufficient attention to empirical data.

  22. Neil Rickert: It has no implications for finite logic, nor for empirical data.

    Did you read the article?

    quote:

    For a finite chunk of 2D lattice, however, the computation always ends in a finite time, leading to a definite answer. At first sight, therefore, the result would seem to have little relation to the real world. Real materials are always finite, and their properties can be measured experimentally or simulated by computer.

    But the undecidability ‘at infinity’ means that even if the spectral gap is known for a certain finite-size lattice, it could change abruptly — from gapless to gapped or vice versa — when the size increases, even by just a single extra atom.

    end quote:

    It seems that there are implications for at least some empirical questions

    peace

  23. The anticipation is killing me.

    While we are waiting I would like to ask some advice

    I’ve been compiling a list of some operational definitions that will be of use when OMagain completes his work on the game/tool done.

    1) Should I post them here of should I start an OP?
    2) should I post them now so we can flesh them out or wait till the game is ready to provide context?

    peace

  24. Hey Omagain

    It’s been a couple of weeks, how is it going?

    Is there any thing you need from me?
    Did you get the game to work?
    what did you think?
    Are you able to distinguish a “real” string from a randomized copy and models that are close?

    Peace

  25. Over at UD WD400 said

    quote:

    There are plenty of programs that can be shown to halt or not. The halting problem only tells us that it’s not possible to generate a general solution. Likewise (as I understand it) the physics results tells there is no general solution that will allow us to predict arbitrarily small spectral gaps. That doesn’t mean specific solutions can’t be found.

    end quote:

    The difference between specific and general solutions is a big part of the point of the game.

    peace

  26. OMagain: Making progress.

    Excellent,

    I must admit it part of me is glad it is taking you a little time. It’s such a very simple gizmo that in my mind I imagined it could be whipped out in mere minutes.

    I was feeling especially stupid when in reality it proved difficult for me to put together and share.

    To see that it’s taking the Great OMagain a little time to produce it eases some of the inferiority complex I had ;-). Thanks again for the effort you are putting into this.

    That being said I’m really excited to begin some hypothesis testing. It will be doubly rewarding for me since that folks from both sides of the fence had a hand in the endeavor.

    thanks again

    peace

  27. Found the thread.
    Cool I was getting paranoid.
    I haven’t been so excited since I was a little kid. I’ve been doing lots of thinking about all this and I can’t wait to share.

    I think we might finally have something that will move the discussion forward that both sides will agree on.

    peace

Leave a Reply