Abstract
Let \((B_t)_{t\in [0,\infty )}\) be a real Brownian motion on a probability space \((\varOmega ,{\mathcal {F}},P)\). Our concern is whether and how a noncausal type stochastic differential \(dX_t=a(t,\omega )\,dB_t+b(t,\omega )\,dt\) is determined from its stochastic Fourier coefficients (SFCs for short) \((e_n,dX):\)\(=\int _{0}^L\overline{e_n(t)}\,dX_t\) with respect to a CONS \((e_n)_{n\in {\mathbb {N}}}\) of \(L^2([0,L];{\mathbb {C}})\). This problem was proposed by Ogawa (Stochastics (85)(2), 286–294, 2013) and has been studied by Ogawa and Uemura (Ogawa in Ind J Stat 77-A(1):30–45, 2014, Ind J Stat 80-A:267–279, 2018; Ogawa and Uemura in J Theor Probab 27:370–382, 2014, Bull Sci Math 138:147–163, 2014, RIMS K\({\hat{\mathrm{o}}}\)ky\({\hat{\mathrm{u}}}\)roku 1952:128–134, 2015, J Ind Appl Math 35-1:373–390, 2018). In this paper we give several results on the problem for each of stochastic differentials of Ogawa type and Skorokhod type when [0, L] is a finite or an infinite interval. Specifically, we first give a condition for a random function to be determined from the SFCs and apply it to obtain affirmative answers to the question with several concrete derivation formulas of the random functions.
Similar content being viewed by others
References
Dieudonné, J.: Foundations of Modern Analysis. Academic Press, New York (1969)
Hoshino, K.: Identification of noncausal finite variation processes from the stochastic Fourier coefficients, preprint arXiv:2001.11594 (2020)
Hoshino, K., Kazumi, T.: On the Ogawa integrability of noncausal Wiener functionals. Stochastics 91(5), 773–796 (2019)
Hoshino, K., Kazumi, T.: On the identification of noncausal Wiener functionals from the stochastic Fourier coefficients. J. Theor. Probab. 32, 1973–1989 (2019)
Itô, K., Nisio, M.: On the convergence of sums of independent Banach space valued random variables. Osaka J. Math. 5, 35–48 (1968)
Karatzas, I., Shreve, S.E.: Brownian Motion and Stochastic Calculus, 2nd edn. Springer, New York (1998)
Malliavin, P., Mancino, M.E.: Fourier series method for measurement of multivariate volatilities. Finance Stoch. 6(1), 49–61 (2002)
Malliavin, P., Thalmaier, A.: Stochastic Calculus of Variations in Mathematical Finance. Springer, Berlin (2005)
Ogawa, S.: Sur le produit direct du bruit blanc par lui-même. C.R. Acad. Sci. Paris Sér. A 288, 359–362 (1979)
Ogawa, S.: Noncausal problems and stochastic analysis (in Japanese). RIMS Kôkyûroku 527, 1–25 (1984)
Ogawa, S.: Quelques propriétés de l’intégrale stochastique du type noncausal. Jpn. J. Appl. Math. 1, 405–416 (1984)
Ogawa, S.: On the stochastic integral equation of Fredholm type, waves and patterns. In: Kinokuniya and North-Holland, pp. 597–605 (1986)
Ogawa, S.: Topics in the theory of noncausal stochastic integral equations. In: M. Pinsky (ed.) Diffusion Processes and Related Problems in Analysis, vol. 1. Birkhauser Boston Inc., Boston, pp. 411–420 (1990)
Ogawa, S.: Seminairs de Probabilités—XXV. Stochastic integral equations for the random fields, pp. 324–329. Springer, Berlin (1991)
Ogawa, S.: On a stochastic Fourier transformation. Stochastics 85(2), 286–294 (2013)
Ogawa, S.: A direct inversion formula for SFT, Sankhya. Indian J. Stat. 77–A(1), 30–45 (2014)
Ogawa, S.: Noncausal Stochastic Calculus. Springer, New York (2017)
Ogawa, S.: Direct inversion formulas for the natural SFT, Sankhya. Indian J. Stat. 80–A, 267–279 (2018)
Ogawa, S., Uemura, H.: On a stochastic Fourier coefficient: case of noncausal function. J. Theor. Probab. 27, 370–382 (2014)
Ogawa, S., Uemura, H.: Identification of a noncausal Itô process from the stochastic Fourier coefficients. Bull. Sci. Math. 138, 147–163 (2014)
Ogawa, S., Uemura, H.: On the identification of noncausal functions from the SFCs (symposium on probability theory). RIMS Kôkyûroku 1952, 128–134 (2015)
Ogawa, S., Uemura, H.: Some aspects of strong inversion formulas of an SFT. Jpn. J. Ind. Appl. Math. 35–1, 373–390 (2018)
Skorokhod, A.V.: On a generalization of a stochastic integral. Theory Probab. Appl. 20, 219–233 (1975)
Acknowledgements
The author would like to express his grate gratitude to Professor Tetsuya Kazumi for supervising to draw this paper in English and to Professor Shigeyoshi Ogawa and Professor Hideaki Uemura for paying kind attention and giving me some advice and comments to this study on several occasions.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
There are no conflicts of interest related to this publication and this research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In Appendix, we follow the notation and terminology in Subsection 2.1. The following Propositions 3 and 4 are concerned with measurability and continuity of a stochastic process whose paths are of bounded variation.
Proposition 3
Let\(a:[0,L]\times \varOmega \rightarrow \mathbb {R}\)be a weak stochastic process, namely, \(a_t\)is a random variable for every\(t\in [0,L]\). Suppose all paths ofa(t) are left-continuous and of bounded variation. Then, both\(a_{+}(t)\)and\(a_{-}(t)\)are weak stochastic processes, and their all paths are left-continuous.
Remark
In particular, both \(a_{+}(t)\) and \(a_{-}(t)\) become measurable stochastic process since any weak stochastic process whose all paths are left-continuous is measurable (see Proposition 1.13 in [6] for example).
Proof
First, the left-continuity of \(a_{+}(t)\) and \(a_{-}(t)\) follows from the left-continuity of a(t). Next, we show that \(a_{-}(t)\) is a weak stochastic process. We can also show so is \(a_+(t)\) in the same manner. Fix any \(t\in (0,L]\). Put \(V(t_0,...,t_n)=\textstyle \sum \limits \nolimits _{j=1}^{n}(a(t_j)-a(t_{j-1}))^{-}\) for \(n\in \mathbb {N}\) and \(t_0,\ldots ,t_n\) such that \(0=t_0<t_1<\cdots <t_n=t\). All of them are random variables since a(t) is the weak stochastic process. While \(a_{-}(t)\) is represented as
setting \(A=\{\,(t_0,...,t_n)\,|\,n\in \mathbb {N},\, 0=t_0<t_1<\cdots <t_n=t,\, t_1,...,t_{n-1}\in \mathbb {Q}\,\}\), (35) is rephrased as
by the denseness of \(\mathbb {Q}\) and the left-continuity of a(t). Besides, the cardinality of A equals that of \(\displaystyle {{\,{\cup }\,}}_{n=1}^{\infty }\mathbb {Q}^{n-1}\) (\(\mathbb {Q}^{0}\) denotes some singleton), which is a countable set. Therefore, \(a_{-}(t)\) is a random variable. \(\square\)
Proposition 4
Leta(t) be a noncausal finite variation process on [0, L]. Then, there exists a version\(\widetilde{a}(t)\)ofa(t) in\(L^0([0,L]\times \varOmega )\)such that
- 1.
All paths of\(\,\widetilde{a}(t)\)are left-continuous and of bounded variation,
- 2.
\(\widetilde{a}_{+}(t)\)and\(\widetilde{a}_{-}(t)\)are\({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable and all their paths are left-continuous and monotonically increasing.
Proof
By the assumption on a(t), there exists \(\varOmega ^V\in {\mathcal {F}}\) such that \(P(\varOmega ^V)=1\) and a(t) is of bounded variation on \(\varOmega ^V\). Set \(a^{(0)}=a\mathsf {1}_{\varOmega ^V}\), then \(a^{(0)}(t)\) is a version of a(t) in \(L^0([0,L]\times \varOmega )\) and all paths of \(a^{(0)}(t)\) are of bounded variation. Furthermore, set \(\widetilde{a}(t,\omega )=\textstyle \lim \limits _{s\nearrow t}a^{(0)}(s,\omega )\), which is \({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable. Since discontinuous points of a function of bounded variation are at most countable, \(\widetilde{a}(t)\) is a version of \(a^{(0)}(t)\) in \(L^0([0,L]\times \varOmega )\) and all paths of \(\,\widetilde{a}(t)\) are left-continuous and of bounded variation. Therefore by Proposition 3, \(\widetilde{a}_{+}(t)\) and \(\widetilde{a}_{-}(t)\) are measurable stochastic processes whose all paths are left-continuous and monotonically increasing. \(\square\)
In what follows, using Proposition 3, we show that
is a measurable set of \([0,L]\times \varOmega\) for a (noncausal) stochastic process \(a:[0,L]\times \varOmega \rightarrow \mathbb {R}\) whose all paths are left-continuous and of bounded variation.
Proposition 5
Let\(a:[0,L]\times \varOmega \rightarrow \mathbb {R}\) be \({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable. Assume all paths ofa(t) are bounded and monotonically increasing. Then, all Dini derivatives ofa(t)
are\({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable.
Proof
Put \(g_h(t,\omega )=\textstyle \inf \limits _{0<|\delta |<|h|}\frac{a(t+\delta ,\omega )-a(t,\omega )}{\delta }\) for each \(h\in (-1,1)\). We show the measurability of \(D_{+}a(t,\omega )\) as the others are shown by similar arguments. Since \(D_{+}a(t,\omega ) =\textstyle \lim \limits _{n\rightarrow \infty }g_{\frac{1}{n}}(t,\omega ),\) we have only to show for any \(h>0\,\)\(g_h\) is \({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable, in other words, \(g_h^{-1}(\{-\infty \}\cup (-\infty ,c))\in {\mathcal {L}}([0,L])\otimes {\mathcal {F}}\) for any real number c. We first note that
Now, we show that for any \((t,\omega )\in [0,L]\times \varOmega\)
Assume the antecedent and fix one of the \(\delta \in (0,h)\). For the denseness of \(\mathbb {Q}\) there exists \(r\in (0,\delta )\cap \mathbb {Q}\) such that \(\frac{a(t+\delta ,\omega )-a(t,\omega )}{r}<c\). Here, because \(a(\cdot ,\omega )\) monotonically increase we have \(\frac{a(t+r,\omega )-a(t,\omega )}{r}\le \frac{a(t+\delta ,\omega )-a(t,\omega )}{r}<c\). Thus, the consequent holds. Also, the converse of (36) is trivial, so we have the following:
\(\frac{a(\cdot +r)-a(\cdot )}{r}\) is \({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable function for r since \(a(\cdot )\) is \({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable, then \(g_h^{-1}(\{-\infty \}\cup (-\infty ,c))\in {\mathcal {L}}([0,L])\otimes {\mathcal {F}}\) is proved. \(\square\)
Proposition 6
Let\(a:[0,L]\times \varOmega \rightarrow \mathbb {R}\)be a (noncausal) stochastic process whose all paths are left-continuous and of bounded variation. Then, \(S=\{(t,\omega )\in [0,L]\times \varOmega \,|\,\)\(a_{+}(\cdot ,\omega ) \text{ and } a_{-}(\cdot ,\omega ) \text{ are } \text{ differentiable }\)\(\text{ in } t\}\)is a measurable set of\([0,L]\times \varOmega\). In particular, the following are equivalent.
- (i)
\(a(\cdot ,\omega )\)and\(a_\mathrm{tv}(\cdot ,\omega ) \text{ are } \text{ differentiable } \text{ in } t \, \text{ a.a. } \,\, (t,\omega )\in [0,L]\times \varOmega\).
- (ii)
\(( a(\cdot ,\omega )\)and\(a_\mathrm{tv}(\cdot ,\omega ) \text{ are } \text{ differentiable } \text{ in } t \, \text{ a.a. } \,\, t\in [0,L] ) \, \text{ a.a. } \,\, \omega \in \varOmega\).
- (iii)
\(( a(\cdot ,\omega )\)and\(a_\mathrm{tv}(\cdot ,\omega ) \text{ are } \text{ differentiable } \text{ in } t \, \text{ a.a. } \,\, \omega \in \varOmega ) \, \text{ a.a. } \,\, t\in [0,L]\).
Proof
For convenience, we denote \(a_{+}\) and \(a_{-}\) by \(a_1\) and \(a_2\), respectively. For \(i\in \{1,2\}\) set
Then, by Proposition 3 each \(a_i\) turns out to be \({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\)-measurable, thus by Proposition 5, \(S=\displaystyle \displaystyle {{\,{\cap }\,}}_{i=1,2}\{\,(t,\omega )\in [0,L]\times \varOmega \,|\, D^{+}a_i(t,\omega )\le D_{-}a_i(t,\omega ) \text{ and } D^{-}a_i(t,\omega )\le D_{+}a_i(t,\omega ) \text{ and }\)\(|D^{+}a_i(t,\omega )|<\infty \,\}\) belongs to \({\mathcal {L}}([0,L])\otimes {\mathcal {F}}\). The latter statement holds for \(S=\{(t,\omega )\in [0,L]\times \varOmega \,|\,a(\cdot ,\omega ) \text{ and }\)\(a_\mathrm{tv}(\cdot ,\omega ) \text{ are } \text{ differentiable } \text{ in } t \}.\)\(\square\)
About this article
Cite this article
Hoshino, K. Derivation formulas of noncausal finite variation processes from the stochastic Fourier coefficients. Japan J. Indust. Appl. Math. 37, 527–564 (2020). https://doi.org/10.1007/s13160-020-00404-4
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13160-020-00404-4