Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Computability Julia sets Mark Braverman Microsoft Research January 6, 2009.

Similar presentations


Presentation on theme: "1 Computability Julia sets Mark Braverman Microsoft Research January 6, 2009."— Presentation transcript:

1 1 Computability Julia sets Mark Braverman Microsoft Research January 6, 2009

2 2 Julia Sets Focus on the quadratic family. Let c be a complex parameter. Consider the following discrete-time dynamical system on the complex plane: z ↦ z 2 + c The simplest non-trivial polynomial dynamical system on the complex plane.

3 3 The Julia set The filled Julia set K c is the set of initial z’s for which the orbit does not escape to ∞. The Julia set J c is the boundary of K c : J c = ∂ K c. J c is also the set of points around which the dynamics is unstable.

4 4 Example: c = -0.12 + 0.665 i z ↦ z 2 + c

5 5 Iterating z ↦ z 2 - 0.12 + 0.665 i z ↦ z 2 + c

6 6 Iterating z ↦ z 2 - 0.12 + 0.665 i z ↦ z 2 + c

7 7 Iterating z ↦ z 2 - 0.12 + 0.665 i z ↦ z 2 + c

8 8 Iterating z ↦ z 2 - 0.12 + 0.665 i z ↦ z 2 + c

9 9 Iterating z ↦ z 2 - 0.12 + 0.665 i z ↦ z 2 + c

10 10 Iterating z ↦ z 2 - 0.12 + 0.665 i z ↦ z 2 + c

11 11 The Julia set The filled Julia set K c is the set of initial z’s for which the orbit does not escape to ∞. The Julia set J c is the boundary of K c : J c = ∂ K c. J c is also the set of points around which the dynamics is unstable.

12 12 Example: c ≈ - 0.391 - 0.587 i

13 13 Iterating z ↦ z 2 - 0.391 - 0.587 i z ↦ z 2 + c

14 14 Iterating z ↦ z 2 - 0.391 - 0.587 i “rotation” by an angle θ z ↦ z 2 + c

15 15 Computing K c and J c Given the parameter c as an input, compute K c and J c. The parameter c describes the rule of the dynamics – “its world”.

16 16 Computability model We use the “bit-computability” notion, which accounts for the cost of the operations on a Turing Machine. Dates back to [Turing36], [BanachMazur37]. TM c = −0.46768… + 0.56598…i JcJc

17 17 Input – giving c to TM The input c is given by an oracle φ(m). –On query m the oracle outputs a rational approximation of c within an error of 2 -m. TM is allowed to query c with any finite precision. TM c = −0.46768… + 0.56598…i

18 Input – giving c to TM The input c is given by an oracle φ(m). –On query m the oracle outputs a rational approximation of c within an error of 2 -m. TM is allowed to query c with any finite precision. TM c = −0.46768… + 0.56598…i φ(m) m=2 φ (2) = −0.4+0.5 i

19 Input – giving c to TM The input c is given by an oracle φ(m). –On query m the oracle outputs a rational approximation of c within an error of 2 -m. TM is allowed to query c with any finite precision. TM c = −0.46768… + 0.56598…i φ(m) m=10 φ(10) = −0.468+0.566 i

20 Input – giving c to TM The input c is given by an oracle φ(m). –On query m the oracle outputs a rational approximation of c within an error of 2 -m. TM is allowed to query c with any finite precision. TM c = −0.46768… + 0.56598…i φ(m) m=7 φ(7) = −0.466+0.567 i

21 21 Output Given a precision parameter n, TM needs to output a 2 -n -approximation of J c, which is a “picture” of the set. TM φ(m) JcJc

22 22 Output A 2 -n -approximation of J c, is made of pixels of size ≈ 2 -n. For each pixel, need to decide whether to paint it white or black. JcJc

23 23 JcJc Coloring a Pixel We use round pixels – equivalent up to a constant. A pixel is a circle of radius 2 -n with a rational center. No ??? Yes Put it in if it intersects J c. If twice the pixel does not intersect J c, leave it out. Otherwise, don’t care.

24 Summary TypeEmpirical and prior work New Hyperbolicempirically easy; some shown in poly-time poly-time computable Parabolicempirically computable (exp-time) poly-time computable Siegelempirically computable in many cases some are computable some are not Cremerno useful pictures to datecomputable Filled Julia set K c thought to be tightly linked to J c always computable ?

25 25 Theorem [BY07]: There is an algorithm A that computes a number c such that no machine with access to c can compute J c. In fact, computing J c is as hard as solving the Halting Problem. Under a reasonable conjecture from Complex Dynamics, c can be made poly-time computable. TM c JcJc A ?

26 26 Proving the non-computability theorem Consider the family z ↦ z 2 + e 2 π i θ z. When is there a Siegel disc? Theorem [Brjuno’65]: When the function converges.

27 27 Geometric Meaning of Φ ( θ ) r( θ ) is a measure on the size of the Siegel disc. r( θ ) can be computed from J c and vice versa. Theorem [Yoccoz’88], [Buff,Cheritat’03]: The function Φ ( θ ) + log r( θ ) is continuous. In particular, when Φ ( θ )=∞, r( θ )=0. Theorem [BY’07]: There is an explicit poly-time algorithm that generates a θ such that Φ ( θ ) is as hard to compute as the Halting Problem.

28 c such that J c has a Siegel disc JcJc ? A the rotation angle 2πθ of the Siegel disc r( θ ) – the “size” of the Siegel Disc Continuous: heavy machinery [Yoccoz’88], [Buff,Cheritat’03]. AlgebraicGeometric

29 Summary TypeEmpirical and prior work New Hyperbolicempirically easy; some shown in poly-time poly-time computable Parabolicempirically computable (exp-time) poly-time computable Siegelempirically computable in many cases some are computable some are not Cremerno useful pictures to datecomputable Filled Julia set K c thought to be tightly linked to J c always computable ?

30 30 The map of parameters c disconnected; poly-time connected; poly-time poly- time Non- computable

31 Thank You Computability of Julia Sets, M. Braverman and M. Yampolsky. Springer, 2008.


Download ppt "1 Computability Julia sets Mark Braverman Microsoft Research January 6, 2009."

Similar presentations


Ads by Google