15-382 C OLLECTIVE I NTELLIGENCE β S18 L ECTURE 11: D ISCRETE -T IME D YNAMICAL S YSTEMS 2 I NSTRUCTOR : G IANNI A. D I C ARO
O CCURRENCE OF PERIODIC WINDOWS FOR π > π # At π β π # = 3.57 the map becomes Β§ chaotic and the attractor changes from a finite to an infinite set of points The large window at π β 3.83 contains a Β§ stable period-3 orbit π π¦ = π π¦ 1 β π¦ Γ the logistic map is π¦ 012 = π(π¦ 0 ) Β§ , π¦ 017 = π(π π π¦ 0 ) = π 7 (π¦ 0 ) π¦ 015 = π π π¦ 0 Β§ We are looking for 3-period cycles: every point π in a 3-period cycle repeats Β§ every 3 iterates Γ π must satisfy π = π 7 (π) Γ π is a fixed-point of the π 7 map Β§ Unfortunately, the π 7 map is an 8-degree polynomial, a bit complex to study Β§ 2
O CCURRENCE OF PERIODIC WINDOWS FOR π > π # π = 3.835 , inside 3-period window Β§ Intersections between the graph and the diagonal correspond to the solutions of π 7 π¦ = π¦ Β§ Only the black dots correspond to fixed points, and there are 3 of them, corresponding to the the 3-period cycle The slope of the function, |π : | is greater than 1 for the white dots, and Β§ less than 1 for the black ones Β§ For the other intersections, they correspond to fixed points or 1-period 3
O CCURRENCE OF PERIODIC WINDOWS FOR π > π # π = 3.8 , before 3-period window π = 3.835 , inside 3-period window Β§ The 6 intersections of interest have vanished! Β§ Not anymore periodic behavior For some π between 3.8 and 3.835 the graph is tangent to the diagonal Β§ At this critical value of π , the stable and unstable 3-period cycles coalesce Β§ and annihilate in a tangent bifurcation , that sets the beginning of the periodic window It can be computed analytically that this happens at π = 1 + 6 Β§ 4
O CCURRENCE OF PERIODIC WINDOWS FOR π > π # π = 3.835 , inside 3-period window π > 3.835 Β§ Just after the tangent bifurcation, the slope at black dots (periodic points) is β +1 (a bit less) For increasing values of π , hills and valleys become steeper / deeper Β§ The slope of π 7 at the black dots decreases steadily from β +1 to -1. Β§ When this occurs, a flip bifurcation happens, that causes each of the fixed periodic points to split in two Β§ Γ the 3-period cycle becomes a 6-period cycle ! Β§ β¦ the process iterates, bringing the period doubling cascade! 5
W HAT ABOUT SENSITIVITY TO INITIAL CONDITIONS ? Chaos requires aperiodic orbits + sensitivity to initial conditions ΓΌ Aperiodic orbits arise Β§ Sensitivity to initial conditions? Given an initial condition π¦ = , and an nearby point π¦ = + π = , π = β 0 Β§ π expresses the separation between two near initial conditions Β§ π 0 is the separation after π iterates Β§ If |π 0 | β |π = |π 0B then π is called the Lyapounov exponent (of π¦ = ) Β§ Β§ A positive Lyapounov exponent is a signature of chaos Β§ Applies also to flows (like in the figure) 6
L YAPOUNOV EXPONENTS |π 0 | β |π = |π 0B , taking the logarithm Β§ 2 F G π β 0 ln| F H | Β§ We can observe that π 0 = π 0 π¦ = + π = β π 0 (π¦ = ) Β§ I G J H 1F H KI G (J H ) 2 0 ln |(π 0 ) : π¦ = | , for π = β 0 2 π β 0 ln| | = Β§ F H π 0 is a function of all π¦ M , π = 0, β¦,π Γ expansion by chain rule Β§ 2 (π 0 ) : π¦ M 2 ln | (π 0 ) : π¦ M | 0K2 0 β 0K2 0 ln| β π β | = Β§ MQ= MQ= If the limit for π β β exists, this is the Lyapounov exponent for the orbit Β§ starting in π¦ = π is the same for all points in the basin of attraction of an attractor Β§ For stable fixed points and cycles, π < π Β§ For chaotic attractors, π > π Β§ 7
L YPUNOUV EXPONENTS FOR THE LOGISTIC MAP 8
A NOTHER M AP : H ENON MAP Classical settings for getting a chaotic behavior: π = 1.4, π = 0.3 Fractal dimension β 1.26 9
F ROM N- DIMENSIONAL M APS TO C ELLULAR A UTOMATA Β§ Letβs introduce topological notions and 2 2 , π¦ 0 5 , β¦ .π¦ 0 [ ) π¦ 012 = π 2 (π¦ 0 constraints 5 2 , π¦ 0 5 , β¦. π¦ 0 [ ) π¦ 012 = π 5 (π¦ 0 Β§ Each variable represents the time-evolution β¦.. of a spatial location Β§ Two variables can be (or not) neighbors [ 2 , π¦ 0 5 , β¦ .π¦ 0 [ ) π¦ 012 = π [ (π¦ 0 Β§ Neighboring concepts go beyond metric Β§ Generic k-dimensional map spaces β¦maybe too generic (complex!) Β§ A variableβs evolution only depends on its neighbors Γ Influential variables Β§ Not everybody is neighbor of everybody Cellular Automata (CA) Β§ Γ Each map only depends on a Β§ Mathematical simplification restricted set of neighbors, but the Β§ Modeling spatial relations entire systems stays coupled Β§ Dynamical model of many real-world systems 10
S TATE VARIABLES SPATIALLY BOUNDED ON L ATTICES : C ELLULAR A UTOMATA State variable β State of Spatial location / Cell Β§ Cell in a 1D, or 2D, or 3D Lattice Β§ Example of selected neighborhood of π¦ M , represented by the set { π¦ MK2 , π¦ M12 } π¦ 5 π¦ 7 π¦ [ π¦ 2 π¦ MK2 π¦ M π¦ M12 1D Example of selected neighborhood of π¦ [1M , represented by the set { π¦ [1MK2 , π¦ [1M12 , π¦ M , π¦ M12 ,π¦ MK2 , π¦ 5[1M , π¦ 5[1M12 , π¦ 5[1MK2 } π¦ 5 π¦ 7 π¦ 2 π¦ [ π¦ [12 π¦ [15 π¦ [17 π¦ [1M π¦ 5[ 2D π¦ [1] 11
CA S ARE L ATTICE MODELS Regular π -dimensional discretization of a continuum Β§ E.g., an π -dimensional grid Β§ Β§ Periodic (toroidal) or non periodic structure Β§ More abstract definition: Regular tiling of a space by a primitive cell Bethe lattice, β -connected cycle-free 3D grid lattice, graph where each node is connected filled with spheres to π¨ neighbours, where π¨ is called of different colors the coordination number 12
C ELLULAR A UTOMATA Example of selected neighborhood of π¦ M , represented by the set { π¦ MK2 , π¦ M12 } π¦ 5 π¦ 7 π¦ [ π¦ 2 π¦ MK2 π¦ M π¦ M12 1D 2 2 , π¦ 0 5 ) π¦ 012 = π 2 (π¦ 0 2 2 , π¦ 0 5 , β¦ .π¦ 0 [ ) π¦ 012 = π 2 (π¦ 0 5 2 , π¦ 0 5 , π¦ 0 7 ) π¦ 012 = π 5 (π¦ 0 5 2 , π¦ 0 5 , β¦. π¦ 0 [ ) π¦ 012 = π 5 (π¦ 0 β¦.. β¦.. M , π¦ 0 M MK2 , π¦ 0 M12 ) π¦ 012 = π M (π¦ 0 [ 2 , π¦ 0 5 , β¦ .π¦ 0 [ ) β¦.. π¦ 012 = π [ (π¦ 0 [ [K2 , π¦ 0 [ ) π¦ 012 = π [ (π¦ 0 Toroidal boundary conditions β¦.. 2 [ , π¦ 0 2 , π¦ 0 5 ) π¦ 012 = π 2 (π¦ 0 2 [ , π¦ 0 2 , π¦ 0 5 ) π¦ 012 = π(π¦ 0 β¦.. β¦.. Single map M , π¦ 0 M , π¦ 0 M MK2 , π¦ 0 M12 ) π¦ 012 = π M (π¦ 0 M MK2 , π¦ 0 M12 ) π¦ 012 = π(π¦ 0 β¦.. β¦.. [ [K2 , π¦ 0 [ , π¦ 0 2 ) π¦ 012 = π [ (π¦ 0 [ [K2 , π¦ 0 [ , π¦ 0 2 ) π¦ 012 = π(π¦ 0 13
CA S H ISTORICAL NOTES 14
Recommend
More recommend