Abstract
Based on iteration of random-valued functions we study the problem of solvability in the class of continuous and Hölder continuous functions \(\varphi \) of the equations
where P is a probability measure on a \(\sigma \)-algebra of subsets of \(\Omega \).
Similar content being viewed by others
1 Introduction
Fix a probability space \((\Omega ,{{\mathcal {A}}},P)\), a complete and separable metric space \((X,\rho )\) with the \( \sigma \)-algebra \( {\mathcal {B}} \) of all its Borel subsets, and a \( {\mathcal {B}} \otimes {\mathcal {A}} \)-measurable function \( f: X \times \Omega \rightarrow X \).
We continue the research of continuous solutions \(\varphi :X \rightarrow {\mathbb {R}}\) of the equations
We refer mainly to [2, 5]. Like in these papers we focus on the iteration of random-valued functions:
for \(n \in {\mathbb {N}}\), \(x \in X\) and \( (\omega _1, \omega _2, \ldots ) \) from \( \Omega ^{\infty } \) defined as \( \Omega ^{{\mathbb {N}}} \). Note that for \(n \in {\mathbb {N}}\) the nth iterate \( f^n \) mapping \(X \times \Omega ^{\infty } \) into X is \( {\mathcal {B}} \otimes {\mathcal {A}}_n \)-measurable, where \( {\mathcal {A}}_n \) denotes the \( \sigma \)-algebra of all sets of the form
with A from the product \( \sigma \)-algebra \( {\mathcal {A}}^n \). (See [13, section 1.4], [10].)
Let \( \pi _n^f (x, \cdot ) \) denote the distribution of \( f^n (x, \cdot ) \), i.e.,
If
with a \( \lambda \in (0,1) \), and
then (see [2, Theorem 3.1]) there exists a probability Borel measure \( \pi ^f\) on X such that for every \(x \in X\) the sequence \((\pi _n^f (x,\cdot ))_{n \in {\mathbb {N}}}\) converges weakly to \(\pi ^f\), and (see [11, Corollary 5.6 and Lemma 3.1], also [3, Lemma 2.2]) for every non-expansive \(u:X \rightarrow {\mathbb {R}}\) the inequality
holds for \(x \in X\) and \(n \in {\mathbb {N}}\).
This limit distribution \( \pi ^f\) plays an important role in solving (1) and (2), see [5, Theorem 3.1], [2, Corollary 4.1], [3, Theorem 2.1]. In particular:
(I) If \(F:X \rightarrow {\mathbb {R}}\) is continuous and bounded, then any continuous and bounded solution \(\varphi :X \rightarrow {\mathbb {R}}\) of (1) has the form
if additionally F is Lipschitz, then (6) defines a Lipschitz solution \(\varphi :X \rightarrow {\mathbb {R}}\) of (1).
(II) If \(F:X \rightarrow {\mathbb {R}}\) is continuous and bounded and (2) has a continuous and bounded solution \(\varphi :X \rightarrow {\mathbb {R}}\), then
and any such solution has the form
with a real constant c.
(III) If \(F:X \rightarrow {\mathbb {R}}\) is Lipschitz, then it is integrable for \( \pi ^f\) and (2) has a Lipschitz solution \(\varphi :X \rightarrow {\mathbb {R}}\) if and only if (7) holds.
The limit distribution \( \pi ^f\) and facts cited above will be used in the main part of the paper. A characterization of this limit for some special random-valued functions in Hilbert spaces have been given by [3, Theorem 3.1] and, in Banach spaces, by [4, Theorem 2.1].
Actually we do not have a sufficiently satisfactory theorem to guarantee the existence of continuous solutions to the equations considered. An explanation of this situation is given in the paper [9] by Witold Jarczyk (see also [13, Note 3.8.4]). Namely, in the case where \(\Omega \) is a singleton and X is a compact real interval, for the appropriate f the set of continuous \(F:X \rightarrow {\mathbb {R}}\) such that the equation has a continuous solution is small in the sense of Baire category. It is also small from the measure point of view (see [1]). We will go also in this direction but, above all, we are looking for conditions under which Eqs. (1) and (2) have continuous and Hölder continuous solutions \(\varphi :X \rightarrow {\mathbb {R}}\). In the case where \(\Omega \) is a singleton, see [12, Chapter II, §7].
2 Results
We will consider Eqs. (1) and (2) assuming the following hypothesis (H).
(H) \((\Omega ,{{\mathcal {A}}},P)\) is a probability space, \((X,\rho )\) is a complete and separable metric space, \( f: X \times \Omega \rightarrow X \) is \( {\mathcal {B}} \otimes {\mathcal {A}} \)-measurable, (3) holds with a \( \lambda \in (0,1) \) and (4) is satisfied.
We regard \(\lambda \) as fixed in (0, 1), and for any metric space X we define \({\mathcal {F}}(X)\) as the set of all continuous functions \(F:X \rightarrow {\mathbb {R}}\) such that there are a sequence \((F_n)_{n \in {\mathbb {N}}}\) of real functions on X and constants \(\vartheta \in (0,1)\), \(L \in (0,\frac{1}{\lambda } )\) and \(\alpha , \beta \in (0, \infty )\) such that
and
Clearly any real Lipschitz function defined on X belongs to \({\mathcal {F}}(X)\).
Theorem 2.1
Assume (H). If \(F \in {\mathcal {F}}(X)\), then formula (6) defines a continuous solution \(\varphi :X \rightarrow {\mathbb {R}}\) of (1), and if additionally (7) holds, then the formula
defines a continuous solution \(\varphi _0 :X \rightarrow {\mathbb {R}}\) of (2).
The proof will be based on three lemmas. In each of them we assume (H).
Lemma 2.2
If \(F \in {\mathcal {F}}(X)\), then the integrals
are finite, and the function
is continuous.
Proof
Corresponding to F choose a sequence \((F_n)_{n \in {\mathbb {N}}}\) of real functions on X and constants \(\vartheta \in (0,1)\), \(L \in (0,\frac{1}{\lambda } )\) and \(\alpha , \beta \in (0, \infty )\) as in the definition of \({\mathcal {F}}(X)\). Then
for \(x \in X\), and
see also (III). Moreover, for every \(n \in {\mathbb {N}}\) the function
is Lipschitz:
for \(x,z \in X\), and therefore function (9), as their uniform limit, is continuous.
\(\square \)
Lemma 2.3
If \(F \in {\mathcal {F}}(X)\), then
and for every \(n \in {\mathbb {N}}\) the function
is continuous.
Proof
and
Since
an application of Lemma 2.2 with f replaced by \(f^n, \ n \in {\mathbb {N}},\) finishes the proof.
\(\square \)
Lemma 2.4
If \(F \in {\mathcal {F}}(X)\), then there are constants \(\theta \in (0,1)\) and \(M \in (0,\infty )\) such that
for \(x, x_0 \in X\) and \(n \in {\mathbb {N}}\).
Proof
Corresponding to F choose a sequence \((F_n)_{n \in {\mathbb {N}}}\) of real functions on X and constants \(\vartheta \in (0,1)\), \(L \in (0,\frac{1}{\lambda } )\) and \(\alpha , \beta \in (0, \infty )\) as in the definition of \({\mathcal {F}}(X)\), and put
Then \(\theta \in (0,1)\), and by Lemmas 2.3 and 2.2, (5) with \(u=\frac{F_n}{\beta L^n}\) and (3) for every \(x, x_0 \in X\) and \(n \in {\mathbb {N}}\) we have
\(\square \)
Proof of Theorem 2.1
It follows from Lemmas 2.2–2.4 that formula (6) defines a continuous function \(\varphi :X \rightarrow {\mathbb {R}}\) and arguing like in the proof of Theorem 3.1(ii) of [5] (see also the calculations below) we show that it solves (1).
Assume now that also (7) holds. Then it follows from Lemmas 2.3 and 2.4 that formula (8) defines a continuous function \(\varphi _0 :X \rightarrow {\mathbb {R}}\). Applying (11), Lemma 2.4, the Lebesgue dominated convergence theorem and the Fubini theorem we observe that for every \(x\in X\) the function \(\varphi _0 \circ f(x,\cdot )\) is integrable for P and
\(\square \)
Proposition 2.5
If F is a real function on a metric space X and
with some constants \(\alpha \in (0,1), \ \beta \in [0,\infty )\), then \(F \in {\mathcal {F}}(X)\).
Proof
Fix \(L \in (1,\frac{1}{\lambda })\), put
and for every \(n \in {\mathbb {N}}\) let \(A_n\) be a maximal for inclusion subset of X such that
By the maximality,
If \(n \in {\mathbb {N}}\) and x, z are distinct points of \(A_n\), then by (12),
It follows from this, using Kirszbraun–McShane extension theorem [7, Theorem 6.1.1], that for every \(n \in {\mathbb {N}}\) there exists an \(F_n:X \rightarrow {\mathbb {R}}\) such that
If \(n \in {\mathbb {N}}\) and \(x \in X\), then there is a \(z \in A_n\) such that \(\rho (x,z)<\theta ^n\), and
\(\square \)
Corollary 2.6
Assume (H). If \(F:X \rightarrow {\mathbb {R}}\) satisfies (12) with some constants \(\alpha \in (0,1), \ \beta \in [0,\infty )\), then formula (6) defines a solution \(\varphi :X \rightarrow {\mathbb {R}}\) of (1) such that
and if additionally (7) holds, then formula (8) defines a solution \(\varphi _0 :X \rightarrow {\mathbb {R}}\) of (2) such that
Proof
By Proposition 2.5 and Theorem 2.1 formula (6) defines a solution \(\varphi :X \rightarrow {\mathbb {R}}\) of (1). Using (6), (11), (12), Jensen’s inequality and (10) for every \(x,z \in X\) we have
For the second part we argue similarly. \(\square \)
Regarding the uniqueness of solutions, we have the following theorem.
Theorem 2.7
Assume (H) and let \(F:X \rightarrow {\mathbb {R}}\).
(i) If \(\varphi _1, \varphi _2 \in {\mathcal {F}}(X) \) are solutions of (1), then \(\varphi _1= \varphi _2.\)
(ii) If \(\varphi _1, \varphi _2 \in {\mathcal {F}}(X) \) are solutions of (2), then \(\varphi _1- \varphi _2\) is a constant function.
Proof
Let \(\varphi _1, \varphi _2 \in {\mathcal {F}}(X) \) and put \(\varphi =\varphi _1- \varphi _2\). Then \(\varphi \in {\mathcal {F}}(X) \). Corresponding to \(\varphi \) choose a sequence \((F_n)_{n \in {\mathbb {N}}}\) of real functions on X and constants \(\vartheta \in (0,1)\), \(L \in (0,\frac{1}{\lambda } )\) and \(\alpha , \beta \in (0, \infty )\) as in the definition of \({\mathcal {F}}(X)\).
If \(\varphi _1, \varphi _2\) are solutions of (1), then \(\varphi \) solves (1) with \(F=0\), and, by induction,
If \(\varphi _1, \varphi _2\) are solutions of (2), then \(\varphi \) solves (2) with \(F=0\), and
In both cases
for \(x, z \in X, \ n \in {\mathbb {N}}\). Moreover,
Consequently, applying among others (10),
whence \(\varphi (x)=\varphi (z)\) for \(x, z \in X\), i.e., \(\varphi \) is a constant function. Noting that if a constant \(\varphi \) solves (1) with \(F=0\), then \(\varphi =0\), we end the proof. \(\square \)
We finish with a qualitative result.
Following [6] by Jens Peter Reus Christensen we say that a Borel subset B of an abelian Polish group G is a Haar zero set if there is a probability Borel measure \(\mu \) on G such that \(\mu (B+x)=0\) for every \(x \in G\). See also [8] where measurability in abelian Polish groups related to Christensen’s Haar zero set is studied.
Assume
(H\(_0\)) \((\Omega ,{{\mathcal {A}}},P)\) is a probability space, \((X,\rho )\) is a compact metric space, \( f: X \times \Omega \rightarrow X \) is \( {\mathcal {B}} \otimes {\mathcal {A}} \)-measurable and (3) holds with a \( \lambda \in (0,1) \).
Assuming (H\(_0\)) we have in particular (4):
Moreover one can consider the Banach space C(X) of all continuous real functions on X with the uniform norm and its subspace \(C_f\),
Clearly \(C_f\) is a closed linear subspace of C(X) and (see, e.g., [7, Corollary 11.2.5]) C(X) is a separable Banach space. We have also the following lemma.
Lemma 2.8
Assume (H\(_0\)). If \(F :X \rightarrow {\mathbb {R}}\) is continuous, then so is the function
Proof
Fix \(\varepsilon \in (0,\infty )\) and choose \(\delta \in (0,\infty )\) such that
Then, by (3), for all \(x,z \in X\),
and therefore the discussed function is continuous. \(\square \)
Let
Theorem 2.9
Under the assumptions (H\(_0\)):
(i) \({\mathcal {F}}_1\) is a Borel and dense subset of C(X), and if \({\mathcal {F}}_1 \not = C(X)\), then \({\mathcal {F}}_1\) is of first category in C(X) and a Haar zero subset of C(X).
(ii) \({\mathcal {F}}_2\) is a Borel and dense subset of \(C_f\), and if \({\mathcal {F}}_2 \not = C_f\), then \({\mathcal {F}}_2\) is of first category in \(C_f\) and a Haar zero subset of \(C_f\).
Proof
By Lemma 2.8 the formulas
define self-mappings \(T_1, T_2\) of C(X). Clearly, these operators are linear and continuous. Moreover,
Furthermore, for every \(F \in T_2(C(X))\) Eq. (2) has a continuous solution \(\varphi :X \rightarrow {\mathbb {R}}\). Hence (II) gives \(T_2(C(X)) \subset C_f\), and
Applying now [1, Lemma] we see that \({\mathcal {F}}_1\) is a Borel subset of C(X), and if \({\mathcal {F}}_1 \not = C(X)\), then \({\mathcal {F}}_1\) is of first category in C(X) and a Haar zero subset of C(X), and \({\mathcal {F}}_2\) is a Borel subset of \(C_f\), and if \({\mathcal {F}}_2 \not = C_f\), then \({\mathcal {F}}_2\) is of first category in \(C_f\) and a Haar zero subset of \(C_f\).
Since by (I) the set
is contained in \({\mathcal {F}}_1\) and (see [7, Theorem 11.2.4]) dense in C(X), the set \({\mathcal {F}}_1\) is dense in C(X).
To show that \({\mathcal {F}}_2\) is dense in \(C_f\), fix \(F \in C_f\) and \(\varepsilon \in (0,\infty )\). Choose a Lipschitz \(F_1: X \rightarrow {\mathbb {R}}\) so that \(\Vert F-F_1\Vert <\frac{\varepsilon }{2}\). According to (III), \(F_1-\int _XF_1d\pi ^f \in {\mathcal {F}}_2\). Moreover,
\(\square \)
Remark 2.10
It is possible that (H\(_0\)) holds and \({\mathcal {F}}_1 = C(X), \ {\mathcal {F}}_2 = C_f\).
To see it consider an \({\mathcal {A}} \)-measurable \(\xi :\Omega \rightarrow X\) and let \(f(x,\omega )=\xi (\omega )\) for \((x,\omega ) \in X \times \Omega \). Then \(f^n(x,\omega )=\xi (\omega _n)\) for \((x,\omega ) \in X \times \Omega ^{\infty } \), so \(\pi _n^f(x,B)=P(\xi \in B)=\pi ^f(B)\) for \(n \in {\mathbb {N}}, \ x \in X, \ B \in {\mathcal {B}} \), and \(\int _XFd\pi ^f=\int _\Omega F \circ \xi dP\) for \(F \in C(X)\). Consequently for every \(F \in C(X)\) the function \(F-\frac{1}{2}\int _XFd\pi ^f\) solves (1), and every \(F \in C_f\) solves (2).
References
Baron, K.: A remark on linear functional equations in the indeterminate case. Glasnik Mat. 20(40), 373–376 (1985)
Baron, K.: On the convergence in law of iterates of random-valued functions. Aust. J. Math. Anal. Appl. 6, no. 1, Art. 3 (2009)
Baron, K.: Weak limit of iterates of some random–valued functions and its application. Aequ. Math. 94, 415–425; 427 (Correction) (2020)
Baron, K.: Around the weak limit of iterates of some random-valued functions. Ann. Univ. Budapest. Sect. Comput. 51, 31–37 (2020)
Baron, K., Kapica, R., Morawiec, J.: On Lipschitzian solutions to an inhomogeneous linear iterative equation. Aequ. Math. 90, 77–85 (2016)
Christensen, J.P.R.: On sets of Haar measure zero in abelian Polish groups. Isr. J. Math. 13, 255–260 (1972)
Dudley, R.M.: Real Analysis and Probability. Cambridge Studies in Advanced Mathematics, vol. 74. Cambridge University Press, Cambridge (2002)
Fischer, P., Słodkowski, Z.: Christensen zero sets and measurable convex functions. Proc. Amer. Math. Soc. 79, 449–453 (1980)
Jarczyk, W.: On a set of functional equations having continuous solutions. Glasnik Mat. 17(37), 59–64 (1982)
Kapica, R.: Sequences of iterates of random-valued vector functions and solutions of related equations. Österreich. Akad. Wiss. Math.-Natur. Kl. Sitzungsber. II 213, 113–118 (2005)
Kapica, R.: The geometric rate of convergence of random iteration in the Hutchinson distance. Aequ. Math. 93, 149–160 (2019)
Kuczma, M.: Functional equations in a single variable. Monografie Matematyczne, vol. 48. PWN–Polish Scientific Publishers, Warszawa (1968)
Kuczma, M., Choczewski, B., Ger, R.: Iterative Functional Equations. Encyclopedia of Mathematics and Its Applications, vol. 32. Cambridge University Press, Cambridge (1990)
Acknowledgements
The research was supported by the Institute of Mathematics of the University of Silesia (Iterative Functional Equations and Real Analysis program).
Author information
Authors and Affiliations
Corresponding author
Additional information
Dedicated to Professor Ludwig Reich on his 80th birthday.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Baron, K. Continuous solutions to two iterative functional equations. Aequat. Math. 95, 1157–1168 (2021). https://doi.org/10.1007/s00010-021-00794-x
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00010-021-00794-x
Keywords
- Iterative functional equations
- Continuous and Hölder continuous solutions
- Random-valued functions
- Iterates
- Convergence in law
- Dense sets
- Sets of first category
- Haar zero sets