# American Institute of Mathematical Sciences

• Previous Article
Impact of spatially heterogeneous temperature on the dynamics of dengue epidemics
• DCDS-B Home
• This Issue
• Next Article
A period doubling route to spatiotemporal chaos in a system of Ginzburg-Landau equations for nematic electroconvection
January  2019, 24(1): 297-319. doi: 10.3934/dcdsb.2017144

## Predicting and estimating probability density functions of chaotic systems

 1 Dept. of mathematics & Statistics, University of Missouri-Kansas City, 5100 Rockhill Rd, Kansas City, MO 64110, USA 2 Dept. of Mathematics & Statistics, Concordia University, 1455 de Maisonneuve Blvd West, Montreal, Quebec, Canada H3G 1M8

Received  November 2016 Revised  March 2017 Published  April 2017

Fund Project: The research of Prof. Paweł Góra was supported by NSERC grant. Also, the research of Profs. Noah Rhee and Majid Bani-Yaghoub was supported by UMKC's funding for excellence program.

In the present work, for the first time, we employ Ulam's method to estimate and to predict the existence of the probability density functions of single species populations with chaotic dynamics. In particular, given a chaotic map, we show that Ulam's method generates a sequence of density functions in L1-space that may converge weakly to a function in L1-space. In such a case we show that the limiting function generates an absolutely continuous (w.r.t. the Lebesgue measure) invariant measure (w.r.t. the given chaotic map) and therefore the limiting function is the probability density function of the chaotic map. This fact can be used to determine the existence and estimate the probability density functions of chaotic biological systems.

Citation: Noah H. Rhee, PaweŁ Góra, Majid Bani-Yaghoub. Predicting and estimating probability density functions of chaotic systems. Discrete & Continuous Dynamical Systems - B, 2019, 24 (1) : 297-319. doi: 10.3934/dcdsb.2017144
##### References:

show all references

##### References:
Graphs of the probability density function $f^*$ and its estimation $f_{2^5}$ for Example 2. Since $\{ f_{2^n} \}_{n \ge 1}$ is bounded by 2, by Theorem 4 $\{ f_{2^n} \}_{n \ge 1}$ is weakly pre-compact, and hence there exists a subsequence $\{ f_{2^{n_k}} \}_{n \geq 1}$ of $\{ f_{2^n} \}_{n \geq 1}$ which converges weakly to $f^* \in L^1[a,b]$. In fact, we may expect that $\{ f_{2^n} \}_{n \ge 1}$ converges strongly to $f^*$.
Graphs of the probability density function $f^*$ and its estimation $f_{2^9}$ for Example 3. Again we may expect that $\{ f_{2^n} \}_{n \ge 1}$ converges strongly to $f^*$.
Bifurcation diagram of Smith-Slatkin map for $\alpha = 1.5, \beta = 1, \gamma = 0.1$, as $p$ changes from 5 to 19.
Graph of Smith-Slatkin map for $\alpha = 1.5, \beta = 1, \gamma = 0.1$ and $p=18$ with $x^*=0.977725973073732$, $x_c=0.857817481013884$, $b=\tau(x_c)=1.295956066342802$, and $a=\tau^2(x_c)=0.147708628629491$. Note that $\tau: [a,b] \rightarrow [a,b]$.
Graph of $\frac{p^3-3p^2+5p-3}{6(p^2+1).}$ For $\alpha=1.5$ and $\gamma=0.1$, the condition $\frac\gamma\alpha < \frac{p^3-3p^2+5p-3}{6(p^2+1)}$ in Proposition 3 is fulfilled for all $p > 2.$
Graph of Smith-Slatkin map for $\alpha = 1.5, \beta = 1, \gamma = 0.1$ and $p=8$. Three first iterates of the critical point are shown.
Graph of Smith-Slatkin map for $\alpha = 1.5, \beta = 1, \gamma = 0.1$ and $p=8.5$. Three first iterates of the critical point are shown.
Graph of $f_{2^8}$ for Example 5. There are six peak values, where the first peak value occurs at $x=a$ and the last peak value at $x=b$. Although the closed form of the probability density function $f^*$ is unknown, the numerical simulations strongly show that $\{ f_{2^n} \}_{n \geq 1}$ is bounded by an integrable function. So, by Theorem 4, $\{ f_{2^n} \}_{n \ge 1}$ is weakly pre-compact, and hence there exists a sub-sequence $\{ f_{2^{n_k}} \}_{n \geq 1}$ of $\{ f_{2^n} \}_{n \geq 1}$ which converges weakly to $f^* \in L^1[a,b]$. By Proposition 2 the measure $\mu$ with density function $f^*$ is an absolutely continuous $\tau$-invariant measure. So $f^*$ is the PDF of the chaotic map $\tau$. In fact, we may expect that $\{ f_{2^n} \}_{n \geq 1}$ converges strongly to $f^*$.
Maximum values of $f_{2^n}$
 $n$ $2^n$ maximum values of$f_{2^n}$ $1$ $2$ $1.25$ $2$ $4$ $1.5075$ $3$ $8$ $1.7232$ $4$ $16$ $1.8157$ $5$ $32$ $1.9126$ $6$ $64$ $1.9465$ $7$ $128$ $1.9748$ $8$ $256$ $1.9871$ $9$ $512$ $1.9930$ $10$ $1024$ $1.9964$
 $n$ $2^n$ maximum values of$f_{2^n}$ $1$ $2$ $1.25$ $2$ $4$ $1.5075$ $3$ $8$ $1.7232$ $4$ $16$ $1.8157$ $5$ $32$ $1.9126$ $6$ $64$ $1.9465$ $7$ $128$ $1.9748$ $8$ $256$ $1.9871$ $9$ $512$ $1.9930$ $10$ $1024$ $1.9964$
Peak values of $f_{2^n}$ for $\tau$ in Example 3
 $n$ $2^n$ 1st peak value of $f_{2^n}$ 2nd peak value of $f_{2^n}$ $5$ $32$ $1.2899$ $3.8291$ $6$ $64$ $1.8337$ $5.4724$ $7$ $128$ $2.4475$ $7.3235$ $8$ $256$ $3.5149$ $10.5309$ $9$ $512$ $4.8498$ $14.5401$ $10$ $1024$ $6.9200$ $20.7534$
 $n$ $2^n$ 1st peak value of $f_{2^n}$ 2nd peak value of $f_{2^n}$ $5$ $32$ $1.2899$ $3.8291$ $6$ $64$ $1.8337$ $5.4724$ $7$ $128$ $2.4475$ $7.3235$ $8$ $256$ $3.5149$ $10.5309$ $9$ $512$ $4.8498$ $14.5401$ $10$ $1024$ $6.9200$ $20.7534$
The consecutive ratio of peak values of $f_{2^n}$ in Example 3
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 1st peak value $f_{2^n}/{f_{2^{n-1}}}$ at 2nd peak value $6$ $64$ $1.4216$ $1.4292$ $7$ $128$ $1.3347$ $1.3382$ $8$ $256$ $1.4361$ $1.4380$ $9$ $512$ $1.3798$ $1.3807$ $10$ $1024$ $1.4269$ $1.4273$
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 1st peak value $f_{2^n}/{f_{2^{n-1}}}$ at 2nd peak value $6$ $64$ $1.4216$ $1.4292$ $7$ $128$ $1.3347$ $1.3382$ $8$ $256$ $1.4361$ $1.4380$ $9$ $512$ $1.3798$ $1.3807$ $10$ $1024$ $1.4269$ $1.4273$
Peak values of $f_{2^n}$ for $\tau$ in Example 5
 $n$ $2^n$ The 1st peak values The 2nd peak values $8$ $256$ $9.9001$ $6.5190$ $9$ $512$ $13.987$ $9.0862$ $10$ $1024$ $20.609$ $12.770$ $11$ $2048$ $28.397$ $16.860$
 $n$ $2^n$ The 1st peak values The 2nd peak values $8$ $256$ $9.9001$ $6.5190$ $9$ $512$ $13.987$ $9.0862$ $10$ $1024$ $20.609$ $12.770$ $11$ $2048$ $28.397$ $16.860$
Peak values of $f_{2^n}$ for $\tau$ in Example 5
 $n$ $2^n$ The 3rd peak values The 4th peak values $8$ $256$ $4.2530$ $2.7992$ $9$ $512$ $5.8493$ $3.8154$ $10$ $1024$ $8.1477$ $5.2147$ $11$ $2048$ $10.711$ $6.8142$
 $n$ $2^n$ The 3rd peak values The 4th peak values $8$ $256$ $4.2530$ $2.7992$ $9$ $512$ $5.8493$ $3.8154$ $10$ $1024$ $8.1477$ $5.2147$ $11$ $2048$ $10.711$ $6.8142$
Peak values of $f_{2^n}$ for $\tau$ in Example 5
 $n$ $2^n$ The 5th peak values The 6th peak values $8$ $256$ $2.0871$ $4.1312$ $9$ $512$ $2.8524$ $5.5220$ $10$ $1024$ $3.5875$ $7.8931$ $11$ $2048$ $4.5694$ $10.607$
 $n$ $2^n$ The 5th peak values The 6th peak values $8$ $256$ $2.0871$ $4.1312$ $9$ $512$ $2.8524$ $5.5220$ $10$ $1024$ $3.5875$ $7.8931$ $11$ $2048$ $4.5694$ $10.607$
The consecutive ratio of the first two peak values of $f_{2^n}$ for $\tau$ in Example 5
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 1st peak value $f_{2^n}/{f_{2^{n-1}}}$ at 2nd peak value $9$ $512$ $1.4128$ $1.3938$ $10$ $1024$ $1.4735$ $1.4055$ $11$ $2048$ $1.3779$ $1.3203$
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 1st peak value $f_{2^n}/{f_{2^{n-1}}}$ at 2nd peak value $9$ $512$ $1.4128$ $1.3938$ $10$ $1024$ $1.4735$ $1.4055$ $11$ $2048$ $1.3779$ $1.3203$
The consecutive ratio of the middle two peak values of $f_{2^n}$ for $\tau$ in Example 5
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 3rd peak value $f_{2^n}/{f_{2^{n-1}}}$ at 4th peak value $9$ $512$ $1.3753$ $1.3630$ $10$ $1024$ $1.3929$ $1.3668$ $11$ $2048$ $1.3146$ $1.3067$
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 3rd peak value $f_{2^n}/{f_{2^{n-1}}}$ at 4th peak value $9$ $512$ $1.3753$ $1.3630$ $10$ $1024$ $1.3929$ $1.3668$ $11$ $2048$ $1.3146$ $1.3067$
The consecutive ratio of the last two peak values of $f_{2^n}$ for $\tau$ in Example 5
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 5th peak value $f_{2^n}/{f_{2^{n-1}}}$ at 6th peak value $9$ $512$ $1.3667$ $1.3367$ $10$ $1024$ $1.2577$ $1.4294$ $11$ $2048$ $1.2737$ $1.3439$
 $n$ $2^n$ $f_{2^n}/{f_{2^{n-1}}}$ at 5th peak value $f_{2^n}/{f_{2^{n-1}}}$ at 6th peak value $9$ $512$ $1.3667$ $1.3367$ $10$ $1024$ $1.2577$ $1.4294$ $11$ $2048$ $1.2737$ $1.3439$
The estimates of ${\bar X}$ and $\sigma$ for Example 5
 $N$ ${\bar x}$ s $10^5$ $0.6658$ $0.3570$ $10^6$ $0.6657$ $0.3574$ $10^7$ $0.6657$ $0.3575$
 $N$ ${\bar x}$ s $10^5$ $0.6658$ $0.3570$ $10^6$ $0.6657$ $0.3574$ $10^7$ $0.6657$ $0.3575$
 [1] Jawad Al-Khal, Henk Bruin, Michael Jakobson. New examples of S-unimodal maps with a sigma-finite absolutely continuous invariant measure. Discrete & Continuous Dynamical Systems - A, 2008, 22 (1&2) : 35-61. doi: 10.3934/dcds.2008.22.35 [2] Simon Lloyd, Edson Vargas. Critical covering maps without absolutely continuous invariant probability measure. Discrete & Continuous Dynamical Systems - A, 2019, 39 (5) : 2393-2412. doi: 10.3934/dcds.2019101 [3] Antonio Giorgilli, Simone Paleari, Tiziano Penati. Local chaotic behaviour in the Fermi-Pasta-Ulam system. Discrete & Continuous Dynamical Systems - B, 2005, 5 (4) : 991-1004. doi: 10.3934/dcdsb.2005.5.991 [4] Jiu Ding, Noah H. Rhee. A unified maximum entropy method via spline functions for Frobenius-Perron operators. Numerical Algebra, Control & Optimization, 2013, 3 (2) : 235-245. doi: 10.3934/naco.2013.3.235 [5] Stefan Klus, Péter Koltai, Christof Schütte. On the numerical approximation of the Perron-Frobenius and Koopman operator. Journal of Computational Dynamics, 2016, 3 (1) : 51-79. doi: 10.3934/jcd.2016003 [6] Gábor Domokos, Domokos Szász. Ulam's scheme revisited: digital modeling of chaotic attractors via micro-perturbations. Discrete & Continuous Dynamical Systems - A, 2003, 9 (4) : 859-876. doi: 10.3934/dcds.2003.9.859 [7] Jie Li. How chaotic is an almost mean equicontinuous system?. Discrete & Continuous Dynamical Systems - A, 2018, 38 (9) : 4727-4744. doi: 10.3934/dcds.2018208 [8] Ruwu Xiao, Geng Li, Yuping Zhao. On the design of full duplex wireless system with chaotic sequences. Discrete & Continuous Dynamical Systems - S, 2019, 12 (4&5) : 783-793. doi: 10.3934/dcdss.2019052 [9] Oliver Jenkinson. Every ergodic measure is uniquely maximizing. Discrete & Continuous Dynamical Systems - A, 2006, 16 (2) : 383-392. doi: 10.3934/dcds.2006.16.383 [10] Stefano Galatolo. Global and local complexity in weakly chaotic dynamical systems. Discrete & Continuous Dynamical Systems - A, 2003, 9 (6) : 1607-1624. doi: 10.3934/dcds.2003.9.1607 [11] Todd Young. Asymptotic measures and distributions of Birkhoff averages with respect to Lebesgue measure. Discrete & Continuous Dynamical Systems - A, 2003, 9 (2) : 359-378. doi: 10.3934/dcds.2003.9.359 [12] Stefan Klus, Christof Schütte. Towards tensor-based methods for the numerical approximation of the Perron--Frobenius and Koopman operator. Journal of Computational Dynamics, 2016, 3 (2) : 139-161. doi: 10.3934/jcd.2016007 [13] José Antonio Carrillo, Marco Di Francesco, Antonio Esposito, Simone Fagioli, Markus Schmidtchen. Measure solutions to a system of continuity equations driven by Newtonian nonlocal interactions. Discrete & Continuous Dynamical Systems - A, 2020, 40 (2) : 1191-1231. doi: 10.3934/dcds.2020075 [14] Marc Kesseböhmer, Sabrina Kombrink. A complex Ruelle-Perron-Frobenius theorem for infinite Markov shifts with applications to renewal theory. Discrete & Continuous Dynamical Systems - S, 2017, 10 (2) : 335-352. doi: 10.3934/dcdss.2017016 [15] Christopher Bose, Rua Murray. The exact rate of approximation in Ulam's method. Discrete & Continuous Dynamical Systems - A, 2001, 7 (1) : 219-235. doi: 10.3934/dcds.2001.7.219 [16] Jacques Féjoz. On "Arnold's theorem" on the stability of the solar system. Discrete & Continuous Dynamical Systems - A, 2013, 33 (8) : 3555-3565. doi: 10.3934/dcds.2013.33.3555 [17] Pengyan Wang, Pengcheng Niu. Liouville's theorem for a fractional elliptic system. Discrete & Continuous Dynamical Systems - A, 2019, 39 (3) : 1545-1558. doi: 10.3934/dcds.2019067 [18] Hongyu He, Naohiro Kato. Equilibrium submanifold for a biological system. Discrete & Continuous Dynamical Systems - S, 2011, 4 (6) : 1429-1441. doi: 10.3934/dcdss.2011.4.1429 [19] Ying-Cheng Lai, Kwangho Park. Noise-sensitive measure for stochastic resonance in biological oscillators. Mathematical Biosciences & Engineering, 2006, 3 (4) : 583-602. doi: 10.3934/mbe.2006.3.583 [20] Jonathan C. Mattingly, Etienne Pardoux. Invariant measure selection by noise. An example. Discrete & Continuous Dynamical Systems - A, 2014, 34 (10) : 4223-4257. doi: 10.3934/dcds.2014.34.4223

2018 Impact Factor: 1.008