Diffusion process with evolution and its parameter estimation

A discrete Markov process in an asymptotic diffusion environment with a uniformly ergodic embedded Markov chain can be approximated by an Ornstein–Uhlenbeck process with evolution. The drift parameter estimation is obtained using the stationarity of the Gaussian limit process. Показано, що дискретни...

Full description

Saved in:
Bibliographic Details
Published in:Кибернетика и системный анализ
Date:2020
Main Authors: Koroliuk, V.S., Koroliouk, D., Dovgyi S.О.
Format: Article
Language:English
Published: Інститут кібернетики ім. В.М. Глушкова НАН України 2020
Subjects:
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/190452
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:Diffusion process with evolution and its parameter estimation / V.S. Koroliuk, D. Koroliouk, S.О. Dovgyi // Кибернетика и системный анализ. — 2020. — Т. 56, № 5. — С. 55–62. — Бібліогр.: 8 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1860236422887243776
author Koroliuk, V.S.
Koroliouk, D.
Dovgyi S.О.
author_facet Koroliuk, V.S.
Koroliouk, D.
Dovgyi S.О.
citation_txt Diffusion process with evolution and its parameter estimation / V.S. Koroliuk, D. Koroliouk, S.О. Dovgyi // Кибернетика и системный анализ. — 2020. — Т. 56, № 5. — С. 55–62. — Бібліогр.: 8 назв. — англ.
collection DSpace DC
container_title Кибернетика и системный анализ
description A discrete Markov process in an asymptotic diffusion environment with a uniformly ergodic embedded Markov chain can be approximated by an Ornstein–Uhlenbeck process with evolution. The drift parameter estimation is obtained using the stationarity of the Gaussian limit process. Показано, що дискретний марковський процес в асимптотичному дифузійному середовищі з рівномірним ергодичним вкладеним ланцюгом Маркова може бути наближений процесом Орнштейна-Уленбека з еволюцією. Оцінку параметра дрейфу отримано з використанням стаціонарності гаусівського граничного процесу. Показано, что дискретный марковский процесс в асимптотической диффузионной среде с равномерной эргодической вложенной цепью Маркова может быть приближен процессом Орнштейна-Уленбека с эволюцией. Оценка параметра дрейфа получена с использованием стационарности гауссовского предельного процесса.
first_indexed 2025-12-07T18:25:10Z
format Article
fulltext UDC 519.24 V.S. KOROLIUK , D. KOROLIOUK, S.Î. DOVGYI DIFFUSION PROCESS WITH EVOLUTION AND ITS PARAMETER ESTIMATION Abstract. A discrete Markov process in an asymptotic diffusion environment with a uniformly ergodic embedded Markov chain can be approximated by an Ornstein–Uhlenbeck process with evolution. The drift parameter estimation is obtained using the stationarity of the Gaussian limit process. Keywords: discrete Markov process, diffusion approximation, asymptotic diffusion environment, Ornstein–Uhlenbeck process, phase merging, drift parameter estimation. We consider a random evolution �( )t , t � 0 , that depends on a random environment Y t( ) , t � 0 , which in turn, is switched by an embedded Markov chain X k , k � 0 . The connection between the continuous-time t � 0 and the discrete-time k � 0 will be explained in the sequel. The purpose of this work is to prove the convergence (in distribution) of the process �( )t , t � 0 , to the Ornstein–Uhlenbeck process under some scaling of the process and its time parameter. The limit will be considered by a small series parameter �� 0 , � � 0 . ASYMPTOTIC DIFFUSION ENVIRONMENT Consider a discrete Markov process in a semi-Markov asymptotic diffusion environment, determined by a solution of the following scaled difference stochastic equation: � � � �� �� � � � � � � �( ) ( ) ( ) ( ) ( )t V Y t Y t n n n n n� �� � � 1 2 1 � , (1) where t nn � �:� 2 , hence t t n n� � � 1 2� � � , n� 0 , �� 0 , for the process increments �� � �� � � � � �( ): ( ) ( )t t t n n n� �� � 1 1 , n � 0 . The asymptotic diffusion environment Yn � , n � 0 , is also a random evolution process generated by a solution of the following scaled difference evolutionary equation: �Y t A Y X A Y X n n n n n n � � � � � �� �( ) ( ; ) ( ; ) ,� � � � 1 0 2 0 , (2) with the embedded Markov chain X X tn n � �: ( )� , n � 0 . The terms A y x0 ( ; ) and A y x( ; ) are Lipschitz functions, together with the first derivative A y xy0 ( ; ) . Here the predictable evolutional component in (1) is determined by the following conditional expectation [1]: V Y t E t Y t tn n n n n n( ) ( ) : [ ( ) | , ( )] ( )� � � � � � � � � �� � � �� �� 1 , where it is assumed that the drift regression function V z( ) is positive: V z( ) � 0 z. ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5 55 © V.S. Koroliuk , D. Koroliouk, S.Î. Dovgyi, 2020 The martingale difference ��� �( )t n�1 , n � 1, generated by the process �� � �( )t n�1 , n � 1, is determined by the following conditional second moment: � � ��� � � � �� � � � � � � � �( ) : [( ( ) ( ) ( )) | ]Y E t V Y t Yn n n n n� 1 2 2 . The embedded Markov chain X X tn n � �: ( )� , t nn � �:� 2 , n � 0 , is supposed to be homogeneous ergodic Markov chain with transition probabilities P x B( , ) , x E� , B �� , having a stationary distribution �( )B , B ��, which satisfies the condition � �( ) ( ) ( , )B dx P x B E � � ; �( )E �1. The stochastic difference equations (1), (2) generate a discrete stochastic basis [2, Ch. 1] with filtration Fm n nY t Y t n m( , ) { ( ), ( ), }� � �� � � � � �� , m � 0 . Now we consider three components ( ( ), ( ), )� � � � � �t Y t Xn n n , n � 0 , as piecewise constant functions with continuous time: � � � � � � � � � � � ( ) ( ) ( ) ( ) ( t t Y t Y t X X n t n n n t n � � � � � � � � � �for 2 1 2)� . In what follows, a solution of equations (1), (2) is given by martingale characterization [3, Section 4.4] of three-component Markov process ( ( ), ( ), )� � �t Y t X t , t � 0 : M t t Y t X Y Xt � � � � �� � � �( ) ( ( ), ( ), ) ( ( ), ( ), )� � �0 0 0 �� L s Y s X ds t s �� � � � �� � 0 2 2[ / ] ( ( ), ( ), ) , and the generator of three-component Markov process ( ( ), ( ), )� � �t Y t X t , t � 0 , is represented as follows [4, Ch. 5]: L c y x E c t Y t X n n n � � � � � �� � � �( , , ) : [ ( ( , ( ), )) |� � � � � 2 1 1 1 � � � � � � �( ) , ( ) , ].t c Y t y X xn n n� � � APPROXIMATION OF A DISCRETE MARKOV PROCESS IN A SYMPTOTIC DIFFUSION ENVIRONMENT Let the singular term A y x0 ( ; ) satisfies the balance condition � E dx A y x� �( ) ( ; )0 0 . (3) The approximation of a discrete Markov process in asymptotic diffusion environment gives the following theorem. Theorem 1. Let the Markov chain X n , n � 0 , be uniformly ergodic with the stationary distribution �( )B , B ��. The finite-dimensional distributions of the discrete Markov process (1), together with asymptotical diffusion Y t� ( ) , t � 0 , converge, as � � 0 , to a diffusion Ornstein–Uhlenbeck process with evolution: ( ( ), ( )) ( ( ), ( )) , ,� � �� �t Y t t Y t t T D� �� � 0 0 0 0 . 56 ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5 The limit two-component diffusion process with evolution ( ( ), ( ))� 0 0t Y t , t � 0 , is set by the generator L0 21 2 ( ) ( , ) ( ) ( , ) ( ) ( , )y c y V y c c y y c yc c� � � �� � � � (4) � � �� ( ) ( , ) � ( ) ( , ) , ( , ),A y c y B y c y C Ry y� � � 1 2 2 0 2 B where by definition � ( ): ( )[ ( ; ) ( ; )]A y dx A y x A y x E � �� � 1 , (5) A y x A y x PR A y xy1 0 0 0( ; ): ( ; ) ( ; )� , (6) � ( ) ( ) ( ; ),B y dx B y x E 2 � � � B y x A y x P R A y x( ; ) ( ; ) [ ] ( ; )� �0 0 0 1 2 � . (7) Here � is the standard identity matrix, P is the transition operator of the Markov chain X t , t � 0 , and the potential kernel R 0 is defined as in [3, Section 5.2]: R P x dx x E 0 1� � � � � �� �( ) , : , ( ): ( ) ( )� �Ï � �� � � � . (8) Remark 1. The limit two-component diffusion process ( ( )� 0 t , Y t0 ( )) , t � 0 , set by the generator (4)–(8), has a stochastic representation by the stochastic differential equation d t V Y t t dt Y t dW t� � �0 0 0 0( ) ( ( )) ( ) ( ( )) ( )� � � , dY t A Y t dt B Y t dW t0 0 0 0( ) � ( ( )) � ( ( )) ( ).� � Consequently, the parameters of the limit diffusion � 0 ( )t , t � 0 , depend on the diffusion process Y t0 ( ) , t � 0 . Proof of Theorem 1. The basic idea is that any Markov process is determined by its generator on the class of real-valued test functions, defined on the set of values of Markov process [5]. First of all, the extended three-component Markov chain is used ( ( ), ( ), ( ) ) , ,� � �� � �t Y t X t X t nn n n n n� � �2 0 , (9) with operator characterization in the following form. Lemma 1. The extended Markov chain is determined by the generator � � � � � �� � �( ) ( , , ) [ ( ) ( ) ] ( , , )x c y x y A x c y x� ��2 � , (10) where the transition operators are defined as follows: � �� � � � �� � � �( ) ( ) : [ ( ( ; ))| ( ) , ( ) , ( )y c E c t y t c Y t y X t x� � � � � ], � � � � �� �( ) ( ) : [ ( ( ; ))| ( ; ) , ( ) ],x y E y Y t x Y t x y X t x� � � �� (11) � � �( ) : ( , ) ( ) .x P x dz z E � � ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5 57 The assertion of Lemma 1 follows from the next argumentation. The extended three-component Markov chain (11), under the additional condition Y t y� ( ) � , X t x� ( ) � , has independent components. So its transition probabilities are given by the product of transition probabilities of each component. An essential step in the proof of Theorem 1 is realized in the next lemma. Lemma 2. The generator (10), (11) of three-component Markov chain (9) on the class of real-valued test functions �( , , )c y x , having bound derivatives up to the third order inclusively, admits an asymptotic representation � � �( ) ( , , )x c y x � (12) � � � � �� �[ ( ) ( ) ( ) ( ; )] ( , , );� � �� 2 1 0 0 � � � � � � �x x y P y x c y x � �0 0( ) ( ) ( ; ) ( ) , ( ) ( ) ( ; ) ( );x y A y x y x y A y x y� � � �� � (13) � 0 21 2 ( ) ( ) ( ) ( ) ( ) ( )y c V y c c y c� � � �� � � . The residual term is expressed as: �� � � �( ; ) ( , , ) , , ( )y x c y x C R� � �0 0 3 2 . Here one intends the uniform convergence for all the arguments. Proof of Lemma 2. We use transformation of generator (12) by the formula � � �( ) ( , , )x c y x � � � � � � ��� �� � � 2 [ ( ( ) ) ( ( ) ) ( ; )] ( , , )� � � � � �A x y y x c y x� . (14) The residual term has the following form: � � � �� � �� �( ; ) ( , , ) ( ( ) )( ( ) ) ( , , )y x c y x y A x c y x� � �� . Then we calculate � � � � � � �� � �� �� � � � �2 2[ ( ) ] ( ) { [ ( ( ; ) ) | ( ; ) ] (� �y c E c t y t y c� c)} � � �[ ( ) ( ; )] ( )� � 0 y y c c� � ; The next term in (17) has the next representation: � � � � �� � �� �� � � � �2 2[ ( ) ] ( ) { [ ( ( ; )) | ( ; ) ] (A x y E y Y t y Y t y y� � y)} � � � ��� � � � �� � � 2 2 21 2 E Y t y y E Y t y y x[ ( ; )] ( ) [ ( ; )] ( ) ( )� � � ( )y � �� � �� � � � � ��[ ( ; ) ( ; )] ( ) ( ; ) ( ) ( )� � � �� 1 0 0 0 21 2 � � � �y x y x y y x y x ( )y , gives the asymptotic expansion in Lemma 2: � � � � � � � � � �( ) ( , , ) [ ( ) ( )x c y x x x� � � �� �2 1 0 � � � 1 2 0 2 0( ( ) ) ( ) ] ( , , ) ( ; ) ( , , )� � � �x y c y x y x c y x� �� . Next, we use the solution of singular perturbation problem for the truncated operator [3]. 58 ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5 Lemma 3. The solution of the singular perturbation problem for the truncated operator is realized on perturbed test functions: � � � � � 0 2 1 0 � �� � �( ) ( , , ) [ ( ) ( )x c y x x A x� � � �� � � � � � 1 2 0 2 0 1 2 2( ( ) ) ( )][ ( , ) ( , , ) ( , , )]� � �x y c y c y x c y x� �� � � � � �L0 ( ) ( , ) ( ) ( , ).y c y x c y� ��� (15) The averaging parameters are determined by the formulas (4)–(8). The limit operator is calculated by the formula L0 0 21 2 ( ) ( , ) ( ) ( , ) � ( ) ( , ) � ( )y c y y c y A y c y B yy y� � � �� � � � ( , )c y . (16) Proof of Lemma 3. To solve the singular perturbation problem for the truncated operator, consider the asymptotic representation by the powers of � : � � � 0 2 1 1 � �� � � � �( ) ( , , ) ( , ) [ ( , , )x c y x c y c y x� � �� � � � � �� � � �0 2 0 1( ) ( , )] [ ( , , ) ( ) ( , , )x c y c y x x c y x� � � � � � �[ ( ) ( ( ) ) ( )] ( , )] ( ) ( , ).� � � � �x x y c y x c y 1 2 0 2 0 � �� Obviously that � �( , )c y � 0 . The balance condition (3) is then used. The solution of the equation � �� �1 0 0( , , ) ( ) ( , )c y x x c y� � is given by the formula [4, Section 5.4]: � �1 0 0( , , ) ( ) ( , )c y x x c y� � � . Lemma 3 implies the following equation � � � �� � �2 0 0( , , ) [ ( ) ( ) ( )] ( , ) ( ) ( , ).c y x x x y c y y c y� � � � L (17) Here, by definition � � � � � �( ) : ( ) ( ) ( ( ) )x x R x x� �0 0 0 0 21 2 . The limit operator is calculated using the balance condition L0 0( ) ( , ) [ ( ) ( )] ( ) ( , ).y c y x x y c y� �� � �� �� � � (18) Recall the projector’s operation: � ��( ) ( ) ( , )x dx B y x E � � � , B y x A y x A y x A y x( , ) ( , ) ( , ) ( ( , ))� �0 0 0 0 21 2 � � . Taking into account the definition of evolutionary operators (13), the limit generator is determined by formula (16). The limit operator (18) provides a solution of equation (17), which is a function of �2 ( , , )c y x . The existence of perturbing functions � i ( )� , i �1 2, , ensures the asymptotical representation (15). That completes the proof of Theorem 1. The volatility is generated by introduction of a random environment Y t0 ( ) , t � 0 , into the diffusion parameter of the Ornstein–Uhlenbeck diffusion process � 0 ( )t , t � 0 . ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5 59 PARAMETER ESTIMATION OF THE LIMIT PROCESS The limit Ornstein–Uhlenbeck diffusion process parameters estimation is substantiated in this section without the assumption of volatility, which greatly changes the kind of estimates. The stationarity of the Gaussian statistical experiment is essentially used [6]. It is known [7], that a diffusion type processes are given by stochastic differential d dt dWt t t t � �( ) . (19) The predictable component satisfies the conditions P dt Tt T t 2 0 1� � �� ! " # $ � � �( ) , , (20) P dtt t 2 0 1 � � � �� ! " # $ �( ) , which ensures the absolute continuity of the measure � � ( ) : { : }B P B� � and the measure � �W B P W B( ) : { : }� � for all B T� = � ( t : 0 t T ) . The Radon–Nicodemus derivative specifies the density of the measure � � � T W d d T( ) : ( , )� , (21) which for processes of diffusion type (19) has the following representation. Theorem 2 [7]. The measure density (21) for processes of diffusion type (19) with additional conditions (20) is given by exponential martingale d d T d dt W t t T t t t T t � � ( , ) exp ( ) ( )� � � �� � ��� �0 2 0 1 2 . (22) In particular, the exponential martingale (21) is determined by a solution of the stochastic Dol�ans–Dade equation [2] d dT T T T� � � ( ) ( ) ( ) , ( )� �0 1 , (23) or in equivalent form: � � T T T Td( ) ( ) ( )� �1 . The relationship of the density (22) with the stochastic Dol�ans–Dade equation (23) can be explained, using the It�o formula for exponential function � ( ) � � � % &exp[ ( ) ( ) ]� � T T 1 2 , with � T t T dt( ): ( )� �0 , % & � �� ( ) : ( )T t T dt2 0 , namely (see [2]): d d d dT T T T T T� � � � � � ( ) ( )[ ( ) ( ) ] ( ) ( )� � % & � % & 1 2 1 2 . Taking into account equality � � � ( ) ( ) ( )T T T� � for exponential function � ( ) , we have a stochastic Dol�ans–Dade differential equation for exponential martingale (22). According to the results of the previous section, the limit diffusion process for normalized discrete Markov processes is the Ornstein–Uhlenbeck process with a linear predictable component d V dt dW t Tt t t �� � � 0 0, , Without limiting of generality, let us put � �1. 60 ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5 The maximum likelihood method for estimating the parameter V0 of a diffusion process with a stochastic differential (19) (� �1) is realized for the logarithm of the measure density (22): L V T V dt V dtT t T t T ( , ) : ln ( )� � � �� �� 0 2 2 02 . Therefore the equation for estimating the maximum likelihood method is max ( , ) / 0 2 0 2 0 0 ' ' � � � �� � V t T T t T L V T V dt V dt , and the estimate of the maximum likelihood method has the following form: V d dtT t T t t T � �� � 0 2 0 / . The least square method estimation of parameter V0 of diffusion process with stochastic differential (19) (� �1) is implemented using equality t T t t T t T td V dt dW 0 0 2 0 0� � �� � � . So we have a relationship V V dW dtT t T t t T 0 0 2 0 � � � � / . The estimation of the least squares method has a representation V d dt T t T t t T0 0 2 0 � � � / . Corollary 1. The estimates of maximum likelihood and least squares coincide: V VT T � 0 . Corollary 2. Estimation by the least squares method, and hence estimation by the method of maximum likelihood of the parameter V0 are strongly consistent: P V V T T 1 0 0lim �� � . (24) Remark 2. In the presence of volatility (see [8]), the maximum likelihood estimate and the least squares estimate are different, but the property of strong consistency (24) is retained. REFERENCES 1. Koroliouk D. Binary statistical experiments with persistent nonlinear regression. Theor. Probability and Math. Statist. 2015. Vol. 91. P. 71–80. 2. Borovskikh Yu.V., Korolyuk V.S. Martingale Approximation. Utrecht: VSP, 1997. 320 p. 3. Ethier S.N., Kurtz T.G. Markov Processes: Characterization and Convergence. New Jersey: Willey, 1986. 534 p. 4. Korolyuk V.S., Limnios N. Stochastic Systems in Merging Phase Space. New Jersey; London: World Scientific. 2005. 331 p. 5. Korolyuk V.S., Koroliouk D. Diffusion approximation of stochastic Markov models with persistent regression. Ukrain. Matem. Journal. 1995. Vol. 47, N 7. P. 928–935. 6. Koroliouk D. Stationary statistical experiments and the optimal estimator for a predictable component. Journal of Mathematical Sciences. 2016. Vol. 214, N 2. P. 220–228. ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5 61 7. Cohen S.N., Elliott R.J. Stochastic Calculus and Applications. Probabilitity and Its Application. Basel: Birkhauser, 2015. 673 p. 8. Bel Hadj Khlifa M., Mishura Yu., Ralchenko K., Shevchenko G., Zili M. Stochastic differential equations with generalized stochastic volatility and statistical estimators. Teoriya Imovirnostei ta Matematychna Statystyka. 2017. Vol. 96. P. 8–20. Íàä³éøëà äî ðåäàêö³¿ 22.03.2020 Â.Ñ. Êîðîëþê , Ä. Êîðîëþê, Ñ.Î. Äîâãèé ÄÈÔÓDzÉÍÈÉ ÏÐÎÖÅÑ Ç ÅÂÎËÞÖ²ªÞ ÒÀ ÎÖ²ÍÞÂÀÍÍß ÉÎÃÎ ÏÀÐÀÌÅÒÐÀ Àíîòàö³ÿ. Ïîêàçàíî, ùî äèñêðåòíèé ìàðêîâñüêèé ïðîöåñ â àñèìïòîòè÷íîìó äèôóç³éíîìó ñåðåäîâèù³ ç ð³âíîì³ðíèì åðãîäè÷íèì âêëàäåíèì ëàíöþãîì Ìàðêîâà ìîæå áóòè íàáëèæåíèé ïðîöåñîì Îðíøòåéíà–Óëåíáåêà ç åâî- ëþö³ºþ. Îö³íêó ïàðàìåòðà äðåéôó îòðèìàíî ç âèêîðèñòàííÿì ñòàö³îíàð- íîñò³ ãàóñ³âñüêîãî ãðàíè÷íîãî ïðîöåñó. Êëþ÷îâ³ ñëîâà: äèñêðåòíèé ìàðêîâñüêèé ïðîöåñ, äèôóç³éíà àïðîêñèìàö³ÿ, àñèìïòîòè÷íå äèôóç³éíå ñåðåäîâèùå, ïðîöåñ Îðíøòåéíà–Óëåíáåêà, ôàçîâå óêðóïíåííÿ, îö³íêà ïàðàìåòðà çñóâó. Â.Ñ. Êîðîëþê , Ä. Êîðîëþê, Ñ.À. Äîâãèé ÄÈÔÔÓÇÈÎÍÍÛÉ ÏÐÎÖÅÑÑ Ñ ÝÂÎËÞÖÈÅÉ È ÎÖÅÍÊÀ ÅÃÎ ÏÀÐÀÌÅÒÐÀ Àííîòàöèÿ. Ïîêàçàíî, ÷òî äèñêðåòíûé ìàðêîâñêèé ïðîöåññ â àñèìïòîòè÷åñ- êîé äèôôóçèîííîé ñðåäå ñ ðàâíîìåðíîé ýðãîäè÷åñêîé âëîæåííîé öåïüþ Ìàðêîâà ìîæåò áûòü ïðèáëèæåí ïðîöåññîì Îðíøòåéíà–Óëåíáåêà ñ ýâîëþ- öèåé. Îöåíêà ïàðàìåòðà äðåéôà ïîëó÷åíà ñ èñïîëüçîâàíèåì ñòàöèîíàðíîñòè ãàóññîâñêîãî ïðåäåëüíîãî ïðîöåññà. Êëþ÷åâûå ñëîâà: äèñêðåòíûé ìàðêîâñêèé ïðîöåññ, äèôôóçèîíàÿ àïïðîêñè- ìàöèÿ, àñèìïòîòè÷åñêàÿ äèôôóçèîííàÿ ñðåäà, ïðîöåññ Îðíøòåéíà–Óëåíáå- êà, ôàçîâîå óêðóïíåíèå, îöåíêà ïàðàìåòðà ñäâèãà. Koroliuk Volodymyr Semenovich, Dr. of Sciences, Academiciam of NAS Ukraine, Emer. Prof., Advisor to the Directorate, Institute of Mathe- matics of the National Academy of Sciences of Ukraine, Kyiv. Koroliouk Dmytro, Dr. of Sciences, Lead Researcher, Institute of Telecommunications and Global Information Space of the National Academy of Sciences of Ukraine, Kyiv, e-mail: dimitri.koroliouk@ukr.net. Dovgyi Stanislav Îleksiyovich Dr. of Sciences, Academiciam of NAS of Ukraine, Director of Dpt. of the Institute of Telecommunications and Global Information Space of the National Academy of Sciences of Ukraine, Kyiv, e-mail: pryjmalnya@gmail.com. 62 ISSN 1019-5262. Êèáåðíåòèêà è ñèñòåìíûé àíàëèç, 2020, òîì 56, ¹ 5
id nasplib_isofts_kiev_ua-123456789-190452
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1019-5262
language English
last_indexed 2025-12-07T18:25:10Z
publishDate 2020
publisher Інститут кібернетики ім. В.М. Глушкова НАН України
record_format dspace
spelling Koroliuk, V.S.
Koroliouk, D.
Dovgyi S.О.
2023-06-08T15:19:07Z
2023-06-08T15:19:07Z
2020
Diffusion process with evolution and its parameter estimation / V.S. Koroliuk, D. Koroliouk, S.О. Dovgyi // Кибернетика и системный анализ. — 2020. — Т. 56, № 5. — С. 55–62. — Бібліогр.: 8 назв. — англ.
1019-5262
https://nasplib.isofts.kiev.ua/handle/123456789/190452
519.24
A discrete Markov process in an asymptotic diffusion environment with a uniformly ergodic embedded Markov chain can be approximated by an Ornstein–Uhlenbeck process with evolution. The drift parameter estimation is obtained using the stationarity of the Gaussian limit process.
Показано, що дискретний марковський процес в асимптотичному дифузійному середовищі з рівномірним ергодичним вкладеним ланцюгом Маркова може бути наближений процесом Орнштейна-Уленбека з еволюцією. Оцінку параметра дрейфу отримано з використанням стаціонарності гаусівського граничного процесу.
Показано, что дискретный марковский процесс в асимптотической диффузионной среде с равномерной эргодической вложенной цепью Маркова может быть приближен процессом Орнштейна-Уленбека с эволюцией. Оценка параметра дрейфа получена с использованием стационарности гауссовского предельного процесса.
en
Інститут кібернетики ім. В.М. Глушкова НАН України
Кибернетика и системный анализ
Системний аналіз
Diffusion process with evolution and its parameter estimation
Дифузійний процес з еволюцією та оцінювання його параметра
Диффузионный процесс с эволюцией и оценка его параметра
Article
published earlier
spellingShingle Diffusion process with evolution and its parameter estimation
Koroliuk, V.S.
Koroliouk, D.
Dovgyi S.О.
Системний аналіз
title Diffusion process with evolution and its parameter estimation
title_alt Дифузійний процес з еволюцією та оцінювання його параметра
Диффузионный процесс с эволюцией и оценка его параметра
title_full Diffusion process with evolution and its parameter estimation
title_fullStr Diffusion process with evolution and its parameter estimation
title_full_unstemmed Diffusion process with evolution and its parameter estimation
title_short Diffusion process with evolution and its parameter estimation
title_sort diffusion process with evolution and its parameter estimation
topic Системний аналіз
topic_facet Системний аналіз
url https://nasplib.isofts.kiev.ua/handle/123456789/190452
work_keys_str_mv AT koroliukvs diffusionprocesswithevolutionanditsparameterestimation
AT korolioukd diffusionprocesswithevolutionanditsparameterestimation
AT dovgyiso diffusionprocesswithevolutionanditsparameterestimation
AT koroliukvs difuzíiniiproceszevolûcíêûtaocínûvannâiogoparametra
AT korolioukd difuzíiniiproceszevolûcíêûtaocínûvannâiogoparametra
AT dovgyiso difuzíiniiproceszevolûcíêûtaocínûvannâiogoparametra
AT koroliukvs diffuzionnyiprocesssévolûcieiiocenkaegoparametra
AT korolioukd diffuzionnyiprocesssévolûcieiiocenkaegoparametra
AT dovgyiso diffuzionnyiprocesssévolûcieiiocenkaegoparametra