f f x
play

f ( f ( x )) Solving Iterated Functions Using Genetic Programming - PowerPoint PPT Presentation

Oct 25, 2022 •268 likes •406 views

f ( f ( x )) Solving Iterated Functions Using Genetic Programming Michael Schmidt Hod Lipson 2010 HUMIES Competition Iterated Functions Iterated Function: Answer: f ( f ( x )) = x f ( x ) = x f ( f ( x )) = x + 2 f ( x ) = x + 1 f ( f ( x ))

  1. f ( f ( x )) Solving Iterated Functions Using Genetic Programming Michael Schmidt Hod Lipson 2010 HUMIES Competition

  2. Iterated Functions Iterated Function: Answer: f ( f ( x )) = x f ( x ) = x f ( f ( x )) = x + 2 f ( x ) = x + 1 f ( f ( x )) = x 4 f ( x ) = x 2 f ( f ( x )) = ( x 2 + 1) 2 +1 f ( x ) = x 2 + 1 f ( f ( x )) = x 2 – 2 f ( x ) = ? Why is this problem so hard for humans?

  3. Test of Intelligence: f ( f ( x )) = x 2 – 2 This problem has become famous in math and physics circles for requiring deep mathematical insight in order to solve. B. A. Brown, A. R. Brown, and M. F. Shlesinger, "Solutions of Doubly "Mathvn journal problems," in Mathvn. vol. 01/2009 mathvn.org, and Higher Order Iterated Equations," Journal of Statistical Physics, vol. 2009. 110, pp. 1087-1097, 2003. Appeared in mathematical The rumored fastest solver competitions Michael Fisher

  4. The known solution requires deep human insight to solve a special case g ( a 2 g -1 ( x )) = x 2 – 2 Assume f ( f ( x )) = g ( a 2 g -1 ( x )): g (2 θ ) = x 2 – 2, Next assume a 2 = 2 and let θ = g -1 ( x ) : g (2 θ ) = g ( θ ) 2 – 2, x 2 – 2 = g ( θ ) 2 – 2 x = g ( θ ) = 2 cos( θ ), By inspection: x = g ( g -1 ( θ )) = 2 cos( g -1 ( θ ) )     1 x       f ( x ) 2 cos  2 cos  Double angle formula:     2

  5. But there are possibly many solutions f ( f ( x )) = x f ( x ) = x f ( x ) = – x f ( x ) = 1/ x This a dark area of mathematics; Only a few special cases of functional problems have ever been solved. Yet, Genetic Programming can find these solutions easily….

  6. f ( f ( x )) = x 2 – 2 Straightforward application of Symbolic Regression n   1        2 Fitness of a candidate f ( x ) = y f f x i i n  i 1 Solutions iterated twice: x 2 – 2 What is f ( x )? f(f(x)) x

  7. Solved in 81 seconds 0 100 -2 80 Converged Runs [%] Fitness [-error] -4 And 60 Solved Reliably: -6 40 -8 20 50 trials 50 trials -10 0 0 1 2 3 0 1 2 3 10 10 10 10 10 10 10 10 Time [seconds] Time [seconds]

  8. Nearly Perfect Fitness     18 16 . 4916 2 ( 1 . 16871 10 ) x  f ( x )        18 18 2 ( 1 . 16871 10 ) x ( 16 . 4916 2 ( 1 . 16871 10 ) x ) The genetic program is trying to take a limit….  b 2 ax    f ( x ) lim  2   ax b 2 ax a

  9. Exactly Correct Symbolicly      b 2 a f ( x )    f f ( x )      2 a f ( x ) b 2 a f ( x )      b 2 a f ( x )    lim f f ( x ) lim      2     a f ( x ) b 2 a f ( x ) a a 2     lim f f ( x ) x 2   a The solution is symbolicly correct

  10. New Solution Found with Genetic Programming f ( f ( x )) = x 2 – 2  1 2 ax    f ( x ) lim  2   ax 1 2 ax a

  11. Human Competitive: • Long-developed and infamous problem in physics and mathematics • Has required deep human insight into mathematics to solve special cases • No other general method exists

  12. Human Competitive: • Long-developed and infamous problem in physics and mathematics • Has required deep human insight into mathematics to solve special cases • No other general method exists The Best Entry: • Entirely new solution found via GP • Fastest this problem has ever been solved • Potential impact in many fields, where such problems have never been solved before

  13. Conclusions Use GP to Solve Iterated Functions f ( f ( x ))

Recommend

More recommend