topbanner_forum
  *

avatar image

Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length
  • Thursday March 28, 2024, 7:36 pm
  • Proudly celebrating 15+ years online.
  • Donate now to become a lifetime supporting member of the site and get a non-expiring license key for all of our programs.
  • donate

Author Topic: A simple one-parameter equation that can be used to reproduce any plot  (Read 4930 times)

mouser

  • First Author
  • Administrator
  • Joined in 2005
  • *****
  • Posts: 40,896
    • View Profile
    • Mouser's Software Zone on DonationCoder.com
    • Read more about this member.
    • Donate to Member
Pretty neat stuff.  The full short paper PDF is linked on the page below.

In a very surprising paper Steven Piantadosi shows that a simple function of one parameter (θ) can fit any collection of ordered pairs {Xi,Yi} to arbitrary precision. In other words, the same simple function can fit any scatter plot exactly, just by choosing the right θ. The intuition comes from chaos theory. We know from chaos theory that simple functions can produce seemingly random, chaotic behavior and that tiny changes in initial conditions can quickly result in entirely different outcomes (the butterfly effect). What Piantadosi shows is that the space traversed in these functions by changing θ is so thick that you can reverse the procedure to find a function that fits any scatter plot.


xtabber

  • Supporting Member
  • Joined in 2007
  • **
  • Posts: 618
    • View Profile
    • Donate to Member
This is just another formulation of the old nonsense about an infinite number of monkeys typing away forever and producing the complete works of Shakespeare.  It has nothing to do with Chaos Theory.

ConstanceJill

  • Supporting Member
  • Joined in 2012
  • **
  • Posts: 205
    • View Profile
    • Donate to Member
Hmm I'm not sure I understand (the paper looks way too complex for me to even try reading it) so I'll probably say something stupid, but… if the "single parameter" can be as precise/long as we want it to be then,  yeah, of course, you can describe any bitmap picture with a single parameter (which would happen to be the whole file's binary contents written as a single number) ?_?

ewemoa

  • Honorary Member
  • Joined in 2008
  • **
  • Posts: 2,922
    • View Profile
    • Donate to Member
This quote from the page was interesting:

the paper also tells us that Occam’s Razor is wrong. Overfitting is possible with just one parameter and so models with fewer parameters are not necessarily preferable even if they fit the data as well or better than models with more parameters.