l ecture 10 d iscrete t ime d ynamical s ystems 1
play

L ECTURE 10: D ISCRETE -T IME D YNAMICAL S YSTEMS 1 I NSTRUCTOR : G - PowerPoint PPT Presentation

15-382 C OLLECTIVE I NTELLIGENCE S18 L ECTURE 10: D ISCRETE -T IME D YNAMICAL S YSTEMS 1 I NSTRUCTOR : G IANNI A. D I C ARO (M ORE ) G ENERAL DEFINITION OF DYNAMICAL SYSTEMS A dynamical system is a 3-tuple ,, : is a set of all


  1. 15-382 C OLLECTIVE I NTELLIGENCE – S18 L ECTURE 10: D ISCRETE -T IME D YNAMICAL S YSTEMS 1 I NSTRUCTOR : G IANNI A. D I C ARO

  2. (M ORE ) G ENERAL DEFINITION OF DYNAMICAL SYSTEMS A dynamical system is a 3-tuple π‘ˆ,𝑇, Ξ¦ : 𝑇 is a set of all possible states of the dynamical system (the state space) Β§ π‘ˆ is the set of values the time (evolution) parameter can take Β§ Ξ¦ is the evolution function of the dynamical system, that associates to Β§ each π’š ∈ 𝑇 a unique image in 𝑇 depending on the time parameter 𝑒, (not all pairs (𝑒, π’š) are feasible, that requires the subset 𝐸) Ξ¦:𝐸 βŠ† π‘ˆΓ—π‘‡ β†’ 𝑇 Ø Ξ¦ 0, π’š = π’š (the initial condition) Ø Ξ¦ 𝑒 2 ,Ξ¦ 𝑒 3 ,π’š = Ξ¦(𝑒 2 + 𝑒 3 ,π’š) , (property of states) for 𝑒 3 , 𝑒 3 + 𝑒 2 ∈ 𝐽(𝑦) , 𝑒 2 ∈ 𝐽(Ξ¦(𝑒 3 𝑦)) , 𝐽 𝑦 = {𝑒 ∈ π‘ˆ ∢ (𝑒, π’š) ∈ 𝐸} Ø The evolution function Ξ¦ provides the system state (the value ) at time 𝑒 for any initial state π’š Ø 𝛿 ; = {Ξ¦ 𝑒,π’š ∢ 𝑒 ∈ 𝐽 π’š } orbit of the system through π’š , starting in π’š , the set of visited states as a function of time 2

  3. T YPES OF D YNAMICAL SYSTEMS Β§ Informally: A dynamical system defines a deterministic rule that allows to know the current state as a function of past states Given an initial condition π’š < = π’š(0) ∈ 𝑇 , a deterministic trajectory Β§ π’š 𝑒 , 𝑒 ∈ 𝐽(π’š < ) is produced by π‘ˆ, 𝑇,Ξ¦ Β§ States can be β€œanything” mathematically well-behaved that represent situations of interest Continuous time dynamical systems (Flows): π‘ˆ open interval of ℝ , Β§ Ξ¦ continous and differentiable function Γ  Differential equations 𝑦̇ = 𝑔(𝑦 𝑒 βˆ’ 𝜐 ) Delay models Β§ Ordinary Differential Equations E 𝑦̇ = 𝑔 𝑦 𝑒 + ∫ 𝑔 𝑦 𝜐 π‘’πœ Β§ Integro-Differential E F Equations , accounting for history I H I H 3 IE H πœ’ 𝑦, 𝑒 = I; H πœ’ 𝑦,𝑒 Β§ Partial Differential Equations , G H accounting for space and time Discrete-time dynamical systems (Maps): π‘ˆ interval of β„€ , Ξ¦ a function Β§ 𝑦 K = 𝑔(𝑦 KL3 ,𝑦 KL2 ,…, 𝑦 KLN ) , iterated updating 3

  4. F ROM LOCAL RULES TO GLOBAL BEHAVIORS ? π‘’π’š 𝑒𝑒 = 𝑔(π’š,𝑒) Flows Maps 𝑦 K = 𝑔(𝑦 KL3 ,𝑦 KL2 ,…, 𝑦 KLN ) βˆ†π‘’ = 1 , when βˆ†π‘’ β†’ 0 Γ  𝑒𝑒 Γ  Differential eq. Β§ For an infinitesimal time, only the instantaneous variation, the velocity , makes sense Γ  The next state is expressed implicitly, and all the instantaneous variations, local in time, must be integrated in order to obtain the global behavior π’š(𝑒) Β§ Also in maps, the time-local iteration rule is a local description that can give rise to extremely complex global behaviors Β§ Γ  How do we integrate the local description into global behaviors? Β§ Γ  How do we predict global behaviors from the local descriptions? 4

  5. M APS Β§ Where maps can arise from? Β§ Inherently discrete-time processes: looking at populations in terms of generations, epidemics in terms of weeks, economy in terms of quarters or years, traffic models per hour, growth per days, … Β§ Discretization of differential equations : Euler method: π’šΜ‡ = π’ˆ(π’š) Γ  π’š KR3 = π’š K + β„Žπ’ˆ(π’š K ) , …Runge-Kutta,… Β§ Β§ Discretization of algebraic equations: Newton’s method for solving 𝑔 𝑦 = 0 Γ  Expand in Taylor series Β§ near 𝑦 K :𝑔 𝑦 = 𝑔 𝑦 K + 𝑦 βˆ’ 𝑦 K 𝑔 T 𝑦 K + β‹― taking the usual linear approximation: 𝑔 𝑦 β‰ˆ 𝑔 𝑦 K + 𝑦 βˆ’ 𝑦 K 𝑔 T 𝑦 K , equating to 0: 𝑦 KR3 = 𝑦 K βˆ’ 𝑔 𝑦 K /𝑔 T 𝑦 K Β§ Let’s focus on one-dimensional maps Β§ Even in one dimension, iterated maps can produce incredibly complex behaviors, including deterministic chaos! Β§ Later on, we will consider multi-dimensional maps defined over a lattice (spatial grid) Γ  Cellular Automata Γ  even more complex behaviors 5

  6. M APS : SAME TERMINOLOGY AS IN FLOWS 𝑔 𝑦 = 2𝑦 , 𝑔 is a map Β§ The orbit of 𝑦 under the map 𝑔 is the set of points: Β§ ,… } = {𝑦, 𝑔 𝑦 , 𝑔 2 𝑦 ,𝑔 Y 𝑦 ,… } {𝑦,𝑔 𝑦 ,𝑔 𝑔 𝑦 , 𝑔 𝑔 𝑔 𝑦 corresponding to the iterated application of the map Β§ The initial point provides the initial conditions A point 𝑦 βˆ— , such that 𝑔 𝑦 βˆ— = 𝑦 βˆ— is a fixed point, the orbits remain Β§ in 𝑦 βˆ— for all future iterations Cobweb plots for individual orbits 6

  7. F IXED POINTS Β§ Starts at 1.6, converges to 1 Β§ Starts at 1.8, converges to -1 Stability of a fixed point 𝑦 βˆ— ? Fixed points correspond to intersection between graph 𝑔(𝑦) and 𝑦 Β§ General map: 𝑦 KR3 = 𝑔(𝑦 K ) , 𝑔(𝑦 βˆ— ) = 𝑦 βˆ— Β§ Let’s consider a near orbit, 𝑦 K = 𝑦 βˆ— + 𝜁 K : is the orbit attracted or Β§ repelled from 𝑦 βˆ— ? If it’s attracted we can say that 𝑦 βˆ— is stable Does the perturbation 𝜁 K grow or decay with π‘œ ? Β§ By the definition, 𝑦 βˆ— + 𝜁 KR3 = 𝑦 KR3 = 𝑔(𝑦 βˆ— + 𝜁 K ) , and using the Taylor Β§ series expansion about 𝑦 βˆ— : 𝑦 βˆ— + 𝜁 KR3 = 𝑦 KR3 = 𝑔 𝑦 βˆ— + 𝜁 K = 𝑔 𝑦 βˆ— + 𝑔 T 𝑦 βˆ— 𝜁 K + 𝑃(𝜁 K2 ) Given that 𝑔(𝑦 βˆ— ) = 𝑦 βˆ— Γ  𝜁 KR3 = 𝑔 T 𝑦 βˆ— 𝜁 K + 𝑃(𝜁 K2 ) Β§ Β§ If we take the linear approximation Γ  Linearized map: 𝜁 KR3 = 𝑔 T 𝑦 βˆ— 𝜁 K Β§ Eigenvalue / multiplier: πœ‡ = 𝑔 T 𝑦 βˆ— Β§ 7

  8. F IXED POINTS Linearized map: 𝜁 KR3 = 𝑔 T 𝑦 βˆ— 𝜁 K Β§ Solution of linearized map: 𝜁 3 = πœ‡πœ < , 𝜁 2 = πœ‡πœ 3 = πœ‡ 2 𝜁 < ….. 𝜁 K = πœ‡ K 𝜁 < Β§ If πœ‡ = |𝑔 T 𝑦 βˆ— | < 1 Γ  𝜁 K β†’ 0, for π‘œ β†’ ∞, and 𝑦 βˆ— is linearly stable Β§ If πœ‡ = |𝑔 T 𝑦 βˆ— | > 1 Γ  𝜁 K β†’ ∞, for π‘œ β†’ ∞, and 𝑦 βˆ— is unstable Β§ The linear stability holds also for the general map Β§ The marginal case πœ‡ = |𝑔 T 𝑦 βˆ— | = 1 doesn’t allow to draw conclusions. Β§ In this case the quadratic term 𝑃(𝜁 K2 ) determines the stability If πœ‡ =0, then the fixed point is said superstable Β§ 𝑦 KR3 = sin𝑦 K Β§ 𝑦 βˆ— = 0 is a fixed point Β§ πœ‡ = 𝑔 T 𝑦 βˆ— = 1 , marginal case Β§ Β§ Cobweb Γ  It’s stable! Is it global? For all orbits 𝑦 K β†’ 0 ? Β§ For any 𝑦 < , 𝑦 3 ∈ [βˆ’1,1] since |sin𝑦 3 | < 1 Β§ 1 Β§ Γ  From cobweb we can say it’s global 8

  9. A NOTHER EXAMPLE , LIMITING BEHAVIOR 𝑦 KR3 = cos 𝑦 K Β§ Kβ†’l 𝑦 K ? … by iterating the map (e.g., use calculator!), 𝑦 K β†’ 0.739.. lim Β§ Solution of trascendental equation 𝑦 = cos𝑦 Β§ The fixed point 0.739… has πœ‡ < 0 Γ  Damped oscillations Β§ For 0 < πœ‡ < 1 convergence to a stable fixed point is monotonic Β§ 9

  10. L OGISTIC MAP 𝑦 KR3 = 𝑠𝑦 K (1 βˆ’ 𝑦 K ) Β§ 𝑦 K is a dimensionless measure of the population in the π‘œ th generation Β§ and 𝑠 is the intrinsic growth rate (with capacity being limited to 1) Let’s restrict 0 ≀ 𝑠 ≀ 4 Β§ Γ  The map maps 0,1 β†’ 0,1 Let’s fix 𝑠 and study the evolution Β§ Β§ Trivially, for small growth rates, 𝑠 < 1, the population always goes 1 𝑦 K extinct, as 𝑦 K β†’ 0 2 For 1 < 𝑠 < 3 , population grows Β§ and eventually reaches a non-zero steady state Watch out: this a time series! 10

  11. A PATH TO THE CHAOS … 11

  12. β€œR EGULAR ” BEHAVIOR , PERIODIC ATTRACTORS 12

  13. β€œR EGULAR ” BEHAVIOR , PERIODIC ATTRACTORS 13

  14. T RANSITION TO CHAOTIC BEHAVIOR 14

  15. C HAOS : S ENSITIVITY TO INITIAL CONDITIONS 15

  16. P ERIODS IN THE LOGISTIC MAP Β§ Oscillating about the previous steady state, alternating between small and large populations Period-4 cycle Β§ Β§ Period-2 cycle: Oscillation repeats every two iterations, periodic orbit Period-doubling to cycles appears by increasing 𝑠 Β§ Β§ They correspond to bifurcations i n phase diagram Β§ Successive bifurcations come faster and faster! Limiting value 𝑠 K β†’ 𝑠 l = 3.569946 … Β§ Β§ Geometric convergence, in the limit the distance between successive values shrink to a constant: 16

  17. C HAOS … Β§ 𝑠 > 𝑠 l ? Β§ For many values of 𝑠, the sequence never settles down to a fixed point or a periodic orbit Β§ Aperiodic, bounded behavior! 17

  18. O RBIT DIAGRAM What happens for larger 𝑠 ? Sure, more chaos …. Even more interesting things! Β§ Orbit diagram: system’s attractors as a function of 𝑠 Β§ Β§ Construction: Choose a value of 𝑠 Β§ Select a random initial condition 𝑦 < and generate the orbit: lets iterate for ~ 300 cycles to let Β§ the system settle down, then plot the next ~ 300 points from the map iterations Move to an adjacent value of 𝑠 and repeat, sweeping the 𝑠 interval Β§ At 𝑠 β‰ˆ 𝑠 l = 3.57 the map Β§ becomes chaotic and the attractor changes from a finite to an infinite set of points For 𝑠 > 3.57 , mixture of order Β§ and chaos, with periodic windows interspersed between clouds of (chaotic) dots 18

  19. O RBIT DIAGRAM 19

  20. C HAOS AND ORDER 20

  21. C HAOS AND ORDER , S ELF - SIMILARITY The large window at 𝑠 β‰ˆ 3.83 Β§ contains a stable period-3 orbit Β§ Looking at the period-3 window even closer: a copy of the orbit diagram reappears in miniature! Γ  Self-similarity 21

Recommend


More recommend