A Quasi-Sure Approach to the Control of Non-Markovian Stochastic Dierential Equations

Size: px
Start display at page:

Download "A Quasi-Sure Approach to the Control of Non-Markovian Stochastic Dierential Equations"

Transcription

1 A Quasi-Sure Approach o he Conrol of Non-Markovian Sochasic Dierenial Equaions Marcel Nuz Firs version: June 16, This version: March 18, Absrac We sudy sochasic dierenial equaions (SDEs) whose drif and diusion coeciens are pah-dependen and conrolled. We consruc a value process on he canonical pah space, considered simulaneously under a family of singular measures, raher han he usual family of processes indexed by he conrols. This value process is characerized by a second order backward SDE, which can be seen as a non- Markovian analogue of he Hamilon-Jacobi-Bellman parial dierenial equaion. Moreover, our value process yields a generalizaion of he G-expecaion o he conex of SDEs. Keywords Sochasic opimal conrol, non-markovian SDE, second order BSDE, G-expecaion, random G-expecaion, volailiy uncerainy, risk measure AMS 2000 Subjec Classicaions 93E20, 49L20, 60H10, 60G44, 91B30 Acknowledgemens Financial suppor by Swiss Naional Science Foundaion Gran PDFM /1 is graefully acknowledged. The auhor hanks Shige Peng, Mee Soner, Nizar Touzi and Jianfeng Zhang for simulaing discussions and wo anonymous referees for helpful commens. 1 Inroducion We consider a conrolled sochasic dierenial equaion (SDE) of he form X = x + 0 μ(r, X, ν r ) dr + 0 σ(r, X, ν r ) dw r, 0 T, (1.1) where ν is an adaped conrol process, W is a Brownian moion and he Lipschiz-coninuous coeciens μ(r, X, ν r ) and σ(r, X, ν r ) may depend on Dep. of Mahemaics, Columbia Universiy, New York, mnuz@mah.columbia.edu 1

2 he pas rajecory {X s, 0 s r} of he soluion. Denoing by X ν he soluion corresponding o ν, we are ineresed in he sochasic opimal conrol problem V 0 := sup E[ξ(X ν )], (1.2) ν where ξ is a given funcional. The sandard approach o such a non-markovian conrol problem (cf. [9, 10]) is o consider for each conrol ν he associaed value process J ν = ess sup E[ξ(X ν ) F ], (1.3) ν: ν=ν on [0,] where (F ) is he given lraion. The dependence on ν reecs he presence of a forward componen in he opimizaion problem. The siuaion is quie dieren in Markovian opimal conrol (cf. [12]), where one uses a single value funcion which depends on cerain sae variables bu no on a conrol. This is essenial o describe he value funcion by a dierenial equaion, such as he Hamilon-Jacobi-Bellman PDE, which is he main meri of he dynamic programming approach. I is worh noing ha his equaion is always backward in ime. An analogous descripion for (1.3) via backward SDEs (BSDEs, cf. [20]) is available for cerain popular problems such as uiliy maximizaion wih power or exponenial uiliy funcions (e.g., [13, 18]) or drif conrol (e.g., [10]). However, his relies on a very paricular algebraic srucure which allows for a separaion of J ν ino a backward par independen of ν and a forward par depending on ν. In his paper, we consider he problem (1.2) on he canonical space by recasing i as V 0 = sup E P ν [ξ(b)], (1.4) ν where P ν is he disribuion of X ν and B is he canonical process, and we describe is dynamic value by a single value process V = {V (ω)}. Formally, V corresponds o a value funcion in he Markovian sense if we see he whole rajecory of he conrolled sysem as a sae variable. Even hough (1.2) has feaures of coupled forward-backward ype, he value process is dened in a purely backward manner: one may say ha by consrucing V on he whole canonical space, we essenially calculae he value for all possible oucomes of he forward par. An imporan ingredien in he same vein is ha V is dened quasi-surely under he family of muually singular measures {P ν }. Raher han forming a family of processes as in (1.3), he necessary informaion is sored in a single process which is dened on a large par of he probabiliy space; indeed, he process V seen under P ν should be hough of as an analogue of J ν. Clearly, his is a necessary sep o obain a (second order) backward SDE. We remark ha [22] considered he same conrol problem (1.2) and also made a connecion o nonlinear expecaions. However, in [22], he value process was considered only under he given probabiliy measure. 2

3 We rs consider a fairly regular funcional ξ and dene V (ω) as a condiional version of (1.4). Applying and advancing ideas from [28] and [17], regular condiional probabiliy disribuions are used o dene V (ω) for every ω and prove a pahwise dynamic programming principle (Theorem 3.2). In a second sep, we enlarge he class of funcionals ξ o an L 1 -ype space and prove ha he value process admis a (quasi-sure) càdlàg modicaion (Theorem 5.1). We also show ha he value process falls ino he class of sublinear expecaions sudied in [19]. Indeed, if ξ is considered as a random variable on he canonical space, he mapping ξ V can be seen as a generalizaion of he G-expecaion [23, 24], which, by [7], corresponds o he case μ 0 and σ(r, X, ν r ) = ν r, where he SDE (1.1) degeneraes o a sochasic inegral. Moreover, V can be seen as a varian of he random G-expecaion [17]; cf. Remark 6.5. Finally, we characerize V by a second order backward SDE (2BSDE) in he spiri of [19]; cf. Theorem 6.4. The second order is clearly necessary since he Hamilon-Jacobi-Bellman PDE for he Markovian case is fully nonlinear, while ordinary BSDEs correspond o semilinear equaions. 2BSDEs were inroduced in [5], and in [27] for he non-markovian case. We refer o [27] for he precise relaion beween 2BSDEs in he quasi-sure formulaion and fully nonlinear parabolic PDEs. We remark ha our approach is quie dieren from he (backward) sochasic parial dierenial equaions sudied in [21] for a similar conrol problem (mainly for unconrolled volailiy) and in [14, 15, 2, 3, 4] for so-called pahwise sochasic conrol problems. The relaion o he pahdependen PDEs, inroduced very recenly in [25], is ye o be explored. The remainder of he paper is organized as follows. In Secion 2 we deail he conrolled SDE and is condiional versions, dene he value process for he case when ξ is uniformly coninuous and esablish is regulariy. The pahwise dynamic programming principle is proved in Secion 3. In Secion 4 we exend he value process o a more general class of funcionals ξ and sae is quasi-sure represenaion. The càdlàg modicaion is consruced in Secion 5. In he concluding Secion 6, we provide he Hamilon-Jacobi- Bellman 2BSDE and inerpre he value process as a varian of he random G-expecaion. 2 Consrucion of he Value Funcion In his secion, we rs inroduce he seing and noaion. Then, we dene he value funcion V (ω) for a uniformly coninuous reward funcional ξ and examine he regulariy of V. 3

4 2.1 Noaion We x a consan T > 0 and le Ω := C([0, T ];R d ) be he canonical space of coninuous pahs equipped wih he uniform norm ω T := sup 0 s T ω s, where is he Euclidean norm. We denoe by B he canonical process B (ω) = ω, by P 0 he Wiener measure, and by F = {F } 0 T he (raw) lraion generaed by B. Unless oherwise saed, probabilisic noions requiring a lraion (such as adapedness) refer o F. For any probabiliy measure P on Ω and any (, ω) [0, T ] Ω, we can consruc he corresponding regular condiional probabiliy disribuion P ω ; cf. [29, Theorem 1.3.4]. We recall ha P ω is a probabiliy kernel on F F T ; i.e., P ω is a probabiliy measure on (Ω, F T ) for xed ω and ω P ω (A) is F -measurable for each A F T. Moreover, he expecaion under P ω is he condiional expecaion under P : E P ω [ξ] = E P [ξ F ](ω) P -a.s. whenever ξ is F T -measurable and bounded. Finally, P ω he se of pahs ha coincide wih ω up o, is concenraed on P ω { ω Ω : ω = ω on [0, ] } = 1. (2.1) While P ω is no dened uniquely by hese properies, we choose and x one version for each riple (, ω, P ). Le [0, T ]. We denoe by Ω := {ω C([, T ]; R d ) : ω = 0} he shifed canonical space of pahs saring a he origin. For ω Ω, he shifed pah ω Ω is dened by ωr := ω r ω for r T, so ha Ω = {ω : ω Ω}. Moreover, we denoe by P0 he Wiener measure on Ω and by F = {Fr} r T he (raw) lraion generaed by B, which can be idenied wih he canonical process on Ω. Given wo pahs ω and ω, heir concaenaion a is he (coninuous) pah dened by (ω ω) r := ω r 1 [0,) (r) + (ω + ω r)1 [,T ] (r), 0 r T. Given an F T -measurable random variable ξ on Ω and ω Ω, we dene he condiioned random variable ξ,ω on Ω by ξ,ω ( ω) := ξ(ω ω), ω Ω. Noe ha ξ,ω ( ω) = ξ,ω ( ω ); in paricular, ξ,ω can also be seen as a random variable on Ω. Then ω ξ,ω ( ω) is FT -measurable and moreover, ξ,ω depends only on he resricion of ω o [0, ]. We noe ha for an Fprogressively measurable process {X r, r [s, T ]}, he condiioned process {Xr,ω, r [, T ]} is F -progressively measurable. If P is a probabiliy on Ω, he measure P,ω on FT dened by P,ω (A) := P ω (ω A), A F T, where ω A := {ω ω : ω A}, 4

5 is again a probabiliy by (2.1). We hen have E P,ω [ξ,ω ] = E P ω [ξ] = E P [ξ F ](ω) P -a.s. Analogous noaion will be used when ξ is a random variable on Ω s and ω Ω s, where 0 s T. We denoe by Ω s := {ω [s,] : ω Ω s } he resricion of Ω s o [s, ], equipped wih ω [s,] := sup r [s,] ω r. Noe ha Ω s can be idenied wih {ω Ω s : ω r = ω for r [, T ]}. The meaning of Ω is analogous. 2.2 The Conrolled SDE Le U be a nonempy Borel subse of R m for some m N. We consider wo given funcions μ : [0, T ] Ω U R d and σ : [0, T ] Ω U R d d, he drif and diusion coeciens, such ha (, ω) μ(, X(ω), ν (ω)) and (, ω) σ(, X(ω), ν (ω)) are progressively measurable for any coninuous adaped process X and any U-valued progressively measurable process ν. In paricular, μ(, ω, u) and σ(, ω, u) depend only on he pas rajecory {ω r, r [0, ]}, for any u U. Moreover, we assume ha here exiss a consan K > 0 such ha μ(, ω, u) μ(, ω, u) + σ(, ω, u) σ(, ω, u) K ω ω (2.2) for all (, ω, ω, u) [0, T ] Ω Ω U. We denoe by U he se of all U-valued progressively measurable processes ν such ha T 0 μ(r, X, ν r ) dr < and T 0 σ(r, X, ν r ) 2 dr < (2.3) hold pah-by-pah for any coninuous adaped process X. Given ν U, he sochasic dierenial equaion X = x + 0 μ(r, X, ν r ) dr + 0 σ(r, X, ν r ) db r, 0 T under P 0 has a P 0 -a.s. unique srong soluion for any iniial condiion x R d, which we denoe by X(0, x, ν). We shall denoe by he disribuion of X(0, x, ν) on Ω and by P (0, x, ν) := P 0 X(0, x, ν) 1 (2.4) P (0, x, ν) := P 0 ( X(0, x, ν) 0) 1 he disribuion of X(0, x, ν) 0 X(0, x, ν) x; i.e., he soluion which is ranslaed o sar a he origin. Noe ha P (0, x, ν) is concenraed on Ω 0 and can herefore be seen as a probabiliy measure on Ω 0. We shall work under he following nondegeneracy condiion. 5

6 Assumpion 2.1. Throughou his paper, we assume ha F X P 0 F for all X = X(0, x, ν), (2.5) where F X P 0 is he P 0 -augmenaion of he lraion generaed by X and (x, ν) varies over R d U. One can consruc siuaions where Assumpion 2.1 fails. For example, if x = 0, σ 1 and μ(r, X, ν r ) = ν r, hen (2.5) fails for a suiable choice of ν; see, e.g., [11]. The following is a posiive resul which covers many applicaions. Remark 2.2. Le σ be sricly posiive denie and assume ha μ(r, X, ν r ) is a progressively measurable funcional of X and σ(r, X, ν r ). Then (2.5) holds rue. Noe ha he laer assumpion is saised in paricular when μ is unconrolled; i.e., μ(r, ω, u) = μ(r, ω). Proof. Le X = X(0, x, ν). As he quadraic variaion of X, he process σσ (r, X, ν r ) dr is adaped o he lraion generaed by X. In view of our assumpions, i follows ha M := σ(r, X, ν r ) db r = X x μ(r, X, ν ) dr has he same propery. Hence B = σ(r, X, ν r ) 1 dm r is again adaped o he lraion generaed by X. Remark 2.3. For some applicaions, in paricular when he SDE is of geomeric form, requiring (2.5) o hold for all x R d is oo srong. One can insead x he iniial condiion x hroughou he paper, hen i suces o require (2.5) only for ha x. Nex, we inroduce for xed [0, T ] an SDE on [, T ] Ω induced by μ and σ. Of course, he second argumen of μ and σ requires a pah on [0, T ], so ha i is necessary o specify a hisory for he SDE on [0, ]. This role is played by an arbirary pah η Ω. Given η, we dene he condiioned coeciens μ,η : [0, T ] Ω U R d, σ,η : [0, T ] Ω U R d d, μ,η (r, ω, u) := μ(r, η ω, u), σ,η (r, ω, u) := σ(r, η ω, u). (More precisely, hese funcions are dened also when ω is a pah no necessarily saring a he origin, bu clearly heir value a (r, ω, u) depends only 6

7 on ω.) We observe ha he Lipschiz condiion (2.2) is inheried; indeed, μ,η (r, ω, u) μ,η (r, ω, u) + σ,η (r, ω, u) σ,η (r, ω, u) K η ω η ω [,r] = K ω ω [,r] 2K ω ω [,r]. We denoe by U he se of all F -progressively measurable, U-valued processes ν such ha T μ(r, X, ν r ) dr < and T σ(r, X, ν r ) 2 dr < for any coninuous F -adaped process X = {X r, r [0, T ]}. For ν U, he SDE X s = η + s μ,η (r, X, ν r ) dr+ s σ,η (r, X, ν r ) db r, s T under P 0 (2.6) has a unique soluion X(, η, ν) on [, T ]. Similarly as above, we dene P (, η, ν) := P 0 ( X(, η, ν) ) 1 o be he disribuion of X(, η, ν) X(, η, ν) η on Ω. Noe ha his is consisen wih he noaion P (0, x, ν) if x is seen as a consan pah. 2.3 The Value Funcion We can now dene he value funcion for he case when he reward funcional ξ is an elemen of UC b (Ω), he space of bounded uniformly coninuous funcions on Ω. Deniion 2.4. Given [0, T ] and ξ UC b (Ω), we dene he value funcion V (ω) = V (ξ; ω) = sup ν U E P (,ω,ν) [ξ,ω ], (, ω) [0, T ] Ω. (2.7) The funcion ξ is xed hroughou Secions 2 and 3 and hence ofen suppressed in he noaion. In view of he double dependence on ω in (2.7), he measurabiliy of V is no obvious. We have he following regulariy resul. Proposiion 2.5. Le [0, T ] and ξ UC b (Ω). Then V UC b (Ω ) and in paricular V is F -measurable. More precisely, V (ω) V (ω ) ρ( ω ω ) for all ω, ω Ω wih a modulus of coninuiy ρ depending only on ξ, he Lipschiz consan K and he ime horizon T. 7

8 The rs source of regulariy for V is our assumpion ha ξ is uniformly coninuous; he second one is he Lipschiz propery of he SDE. Before saing he proof of he proposiion, we examine he laer aspec in deail. Lemma 2.6. Le ψ UC b (Ω ). There exiss a modulus of coninuiy ρ K,T,ψ, depending only on K, T and he minimal modulus of coninuiy of ψ, such ha E P (,ω,ν) [ψ] E P (,ω,ν) [ψ] ρ K,T,ψ ( ω ω ) for all [0, T ], ν U and ω, ω Ω. Proof. We se E[ ] := E P 0 [ ] o alleviae he noaion. Le ω, ω Ω, se X := X(, ω, ν) and X := X(, ω, ν), and recall ha X = X X = X ω and similarly X = X ω. (i) We begin wih a sandard SDE esimae. Le X = M + A and X = M + A be he semimaringale decomposiions and τ T be a sopping ime such ha M, M, A, A are bounded on [, τ]. Then Iô's formula and he Lipschiz propery (2.2) of σ yield ha E[ M τ M τ 2 ] E τ K 2 E K 2 E σ(r, ω X, ν r ) σ(r, ω τ τ ω X ω X 2 r dr X, νr ) 2 dr ( ω ω + X X [,r] ) 2 dr T 2K 2 T ω ω 2 + 2K 2 E [ X X 2 [,r τ]] dr. Hence, Doob's maximal inequaliy implies ha E [ M M 2 [,τ]] 4E[ M τ M τ 2 ] T 8K 2 T ω ω 2 + 8K 2 E [ X X 2 [,r τ]] dr. Moreover, using he Lipschiz propery (2.2) of μ, we also have ha A s A s s K s μ(r, ω X, ν r ) μ(r, ω ( ω ω + X X [,r] ) dr for all s T and hen Jensen's inequaliy yields ha E [ A A 2 [,τ]] 2K 2 T 2 ω ω 2 + 2K 2 T T X, νr ) dr E [ X X 2 [,r τ]] dr. 8

9 Hence, we have shown ha E [ X X 2 ] T [,τ] C0 ω ω 2 + C 0 E [ X X 2 [,r τ]] dr, where C 0 depends only on K and T, and we conclude by Gronwall's lemma ha E [ X X 2 [,τ]] C ω ω 2, C := C 0 e C 0T. By he coninuiy of heir sample pahs, here exiss a localizing sequence (τ n ) n 1 of sopping imes such ha M, M, A, A are bounded on [, τ n ] for each n. Therefore, monoone convergence and he previous inequaliy yield ha E [ X X 2 [,T ]] C ω ω 2. (2.8) (ii) Le ρ be he minimal (nondecreasing) modulus of coninuiy for ψ, ρ(z) := sup { ψ( ω) ψ( ω ) : ω, ω Ω, ω ω [,T ] z }, and le ρ be he concave hull of ρ. Then ρ is a bounded coninuous funcion saisfying ρ(0) = 0 and ρ ρ. Le P := P (, ω, ν) and P := P (, ω, ν), hen P and P are he disribuions of X and X, respecively; herefore, E P [ψ] E P [ψ] = E[ψ(X ) ψ( X )] E [ ρ ( X X [,T ] )]. (2.9) Moreover, Jensen's inequaliy and (2.8) yield ha E [ ρ ( X X [,T ] )] ρ ( E [ X X [,T ] ]) ρ ( E [ X X 2 ] 1/2 ) [,T ] ρ ( ) C ω ω for every n. In view of (2.9), we have E P [ψ] E P [ψ] ρ( C ω ω ); i.e., he resul holds for ρ K,T,ψ (z) := ρ( Cz). Afer hese preparaions, we can prove he coninuiy of V. Proof of Proposiion 2.5. To disenangle he double dependence on ω in (2.7), we rs consider he funcion (η, ω) E P (,η,ν) [ξ,ω ], (η, ω) Ω Ω. Since ξ UC b (Ω), here exiss a modulus of coninuiy ρ (ξ) for ξ; i.e., ξ(ω) ξ(ω ) ρ (ξ) ( ω ω T ), ω, ω Ω. Therefore, we have for all ω Ω ha ξ,ω ( ω) ξ,ω ( ω) = ξ(ω ω) ξ(ω ω) ρ (ξ) ( ω ω ω ω T ) = ρ (ξ) ( ω ω ). (2.10) 9

10 For xed η Ω, i follows ha E P (,η,ν) [ξ,ω ] E P (,η,ν) [ξ,ω ] ρ (ξ) ( ω ω ). (2.11) Fix ω Ω and le ψ ω := ξ,ω. Then ρ (ξ) yields a modulus of coninuiy for ψ ω ; in paricular, his modulus of coninuiy is uniform in ω. Thus Lemma 2.6 implies ha he mapping η E P (,η,ν) [ψ ω ] admis a modulus of coninuiy ρ T,K,ξ depending only on T, K, ξ. In view of (2.11), we conclude ha E P (,η,ν) [ξ,ω ] E P (,η,ν) [ξ,ω ] ρ (ξ) ( ω ω ) + ρ T,K,ξ ( η η ) for all η, η, ω, ω Ω and in paricular ha E P (,ω,ν) [ξ,ω ] E P (,ω,ν) [ξ,ω ] ρ( ω ω ), ρ := ρ (ξ) + ρ T,K,ξ (2.12) for all ω, ω Ω. Passing o he supremum over ν U, we obain ha V (ω) V (ω ) ρ( ω ω ) for all ω, ω Ω, which was he claim. 3 Pahwise Dynamic Programming In his secion, we provide a pahwise dynamic programming principle which is fundamenal for he subsequen secions. As we are working in he weak formulaion (1.4), he argumens used here are similar o, e.g., [26], while [22] gives a relaed consrucion in he srong formulaion (i.e., working only under P 0 ). We assume in his secion ha he following condiional version of Assumpion 2.1 holds rue; however, we shall see laer (Lemma 4.4) ha his exended assumpion holds auomaically ouside cerain nullses. Assumpion 3.1. Throughou Secion 3, we assume ha for all (, η, ν) [0, T ] Ω U. F X P 0 F for X := X(, η, ν), (3.1) The main resul of his secion is he following dynamic programming principle. We shall also provide more general, quasi-sure versions of his resul laer (he nal form being Theorem 5.2). Theorem 3.2. Le 0 s T, ξ UC b (Ω) and se V r ( ) = V r (ξ; ). Then V s (ω) = sup E P (s,ω,ν)[ (V ) s,ω] for all ω Ω. (3.2) ν U s 10

11 We remark ha in view of Proposiion 2.5, we may see ξ V r (ξ; ) as a mapping UC b (Ω) UC b (Ω) and recas (3.2) as he semigroup propery V s = V s V on UC b (Ω) for all 0 s T. (3.3) Some auxiliary resuls are needed for he proof of Theorem 3.2, which is saed a he end of his secion. We sar wih he (well known) observaion ha condiioning he soluion of an SDE yields he soluion of a suiably condiioned SDE. Lemma 3.3. Le 0 s T, ν U s and ω Ω. If X := X(s, ω, ν), hen X,ω = X (, ω s X(ω), ν,ω ) P 0-a.s. for all ω Ω s. Proof. Le ω Ω s. Using he deniion and he ow propery of X, we have X r = ω s + = X + r s r μ s, ω (u, X, ν u ) du + r μ(u, ω s X, νu ) du + σ s, ω (u, X, ν u ) db s u s r σ(u, ω s X, νu ) db s u P s 0 -a.s. for all r [, T ]. Hence, using ha (P0 s),ω = P0 by he P 0 s -independence of he incremens of B s, X,ω r = X,ω + r μ(u, ω X,ω s, νu,ω ) du+ r σ(u, ω s X,ω, ν,ω u ) db u P 0-a.s. (3.4) Since X is adaped, we have X,ω ( ) = X(ω ) = X(ω) on [s, ] and in paricular ω s X,ω = ω s X(ω) X,ω = η X,ω, for η := ω s X(ω). Therefore, recalling ha X s = ω s, (3.4) can be saed as X,ω r = η + r μ,η (u, X,ω, ν,ω ) du + r σ,η (u, X,ω, ν,ω ) db u P 0-a.s.; i.e., X,ω solves he SDE (2.6) for he parameers (, η, ν,ω ). Now he resul follows by he uniqueness of he soluion o his SDE. Given [0, T ] and ω Ω, we dene P(, ω) = { P (, ω, ν) : ν U }. (3.5) These ses have he following invariance propery. 11

12 Lemma 3.4. Le 0 s T and ω Ω. If P P(s, ω), hen P,ω P(, ω s ω) for P -a.e. ω Ω s. Proof. Since P P(s, ω), we have P = P (s, ω, ν) for some ν U s ; i.e., seing X := X(s, ω, ν), P is he disribuion of X s = s μ(r, ω s X, ν r ) dr + s σ(r, ω s X, ν r ) db s r under P s 0. We se ˆμ r := μ(r, ω s X, ν r ) and ˆσ r := σ(r, ω s X, ν r ) and see he above as he inegral s ˆμ r dr + s ˆσ r dbr s raher han an SDE. As in [26, Lemma 2.2], he nondegeneracy assumpion (3.1) implies he exisence of a progressively measurable ransformaion β ν : Ω s Ω s (depending on s, ω, ν) such ha β ν (X s ) = B s P s 0 -a.s. (3.6) Furhermore, a raher edious calculaion as in he proof of [26, Lemma 4.1] shows ha ( 1 P,ω = P0 ˆμ r,βν(ω) dr + ˆσ r,βν(ω) dbr) for P -a.e. ω Ω s. s s Noe ha, abbreviaing ˇω := β ν (ω), we have ˆμ,βν(ω) r = μ ( r, ω s X,ˇω, ν,ˇω r ) = μ ( r, ω s X(ˇω) X,ˇω, ν,ˇω r and similarly for ˆσ,βν(ω). Hence, we deduce by Lemma 3.3 ha In view of (3.6), we have P,ω = P (, ω s X(ˇω), ν,ˇω ) for P -a.e. ω Ω s. X s (β ν (B s )) = B s P -a.s.; (3.7) i.e., X(ˇω) s = X s (ˇω) = ω for P -a.e. ω Ω s, and we conclude ha P,ω = P (, ω s ω, ν,ˇω ) for P -a.e. ω Ω s. (3.8) In paricular, P,ω P(, ω s ω). Lemma 3.5 (Pasing). Le 0 s T, ω Ω, ν U s and se P := P (s, ω, ν), X := X(s, ω, ν). Le (E i ) 0 i N be a nie F s -measurable pariion of Ω s, ν i U for 1 i N and dene ν U s by ν(ω) := 1 [s,) ν(ω) + 1 [,T ] [ν(ω)1 E 0(X(ω) s ) + Then P := P (s, ω, ν) saises P = P on F s and N i=1 ) ] ν i (ω )1 E i(x(ω) s ). P,ω = P (, ω s ω, ν i ) for P -a.e. ω E i, 1 i N. 12

13 Proof. As ν = ν on [s, ), we have X(s, ω, ν) = X on [s, ] and in paricular P = P on F s. Le 1 i N, we show ha P,ω = P (, ω s ω, ν i ) for P -a.e. ω E i. Recall from (3.8) ha P,ω = P (, ω s ω, ν,β ν(ω) ) for P -a.e. ω Ω s, where β ν is dened as in (3.6). Since boh sides of his equaliy depend only on he resricion of ω o [s, ] and X(s, ω, ν) = X on [s, ], we also have ha P,ω = P (, ω s ω, ν,ˇω ) for P -a.e. ω Ω s, (3.9) where ˇω = β ν (ω) is dened as below (3.6). Noe ha by (3.7), ω E i implies X(ˇω) s E i under P. (More precisely, if A Ω s is a se such ha A E i P -a.s., hen {X(ˇω) s : ω A} E i P -a.s.) In fac, since X is adaped and E i F s, we even have ha X(ˇω ω) s E i for all ω Ω, for P -a.e. ω E i. By he deniion of ν, we conclude ha ν,ˇω ( ω) = ν(ˇω ω) = ν i ((ˇω ω) ) = ν i ( ω), ω Ω, for P -a.e. ω E i. In view of (3.9), his yields he claim. Remark 3.6. In [26] and [17], i was possible o use a pasing of measures as follows: in he noaion of Lemma 3.5, i was possible o specify measures P i on Ω, corresponding o cerain admissible conrols, and use he pasing ˆP (A) := P (A E 0 ) + N i=1 EP [ P i (A,ω )1 E i(ω) ] o obain a measure in Ω s which again corresponded o some admissible conrol and saised ˆP,ω = P i for ˆP -a.e. ω E i, (3.10) which was hen used in he proof of he dynamic programming principle. This is no possible in our SDE-driven seing. Indeed, suppose ha ˆP is of he form P (s, ω, ν) for some ω and ν, hen we see from (3.8) ha, when σ is general, ˆP,ω will depend explicily on ω, which conradics (3.10). Therefore, he subsequen proof uses an argumen where (3.10) holds only a one specic ω E i ; on he res of E i, we conne ourselves o conrolling he error. We can now show he dynamic programming principle. Apar from he dierence remarked above, he basic paern of he proof is he same as in [26, Proposiion 4.7]. Proof of Theorem 3.2. Using he noaion (3.5), our claim (3.2) can be saed as sup E P [ ξ s, ω] = sup E P [ (V ) s, ω] for all ω Ω. (3.11) P P(s, ω) P P(s, ω) 13

14 (i) We rs show he inequaliy in (3.11). Fix ω Ω and P P(s, ω). Lemma 3.4 shows ha P,ω P(, ω s ω) for P -a.e. ω Ω s and hence ha E P,ω [ (ξ s, ω ),ω] = E P,ω [ ξ, ω sω ] sup E P [ ξ, ω ] sω P P(, ω sω) = V ( ω s ω) = V s, ω (ω) for P -a.e. ω Ω s. Since V is measurable by Proposiion 2.5, we can ake P (dω)-expecaions on boh sides o obain ha E P [ ξ s, ω] [ = E P E P [,ω (ξ s, ω ),ω]] E P [ V s, ω ]. We ake he supremum over P P(s, ω) on boh sides and obain he claim. (ii) We now show he inequaliy in (3.11). Fix ω Ω, ν U s and le P = P (s, ω, ν). We x ε > 0 and consruc a counable cover of he sae space as follows. Le ˆω Ω s. By he deniion of V ( ω s ˆω), here exiss ν (ˆω) U such ha P (ˆω) := P (, ω s ˆω, ν (ˆω) ) saises V ( ω s ˆω) E P (ˆω) [ξ, ω s ˆω ] + ε. (3.12) Le B(ε, ˆω) Ω s denoe he open [s,] -ball of radius ε around ˆω. Since (Ω s, [s,] ) is a separable (quasi-)meric space and herefore Lindelöf, here exiss a sequence (ˆω i ) i 1 in Ω s such ha he balls B i := B(ε, ˆω i ) form a cover of Ω s. As an [s,] -open se, each B i is F s -measurable and hence E 1 := B 1, E i+1 := B i+1 (E 1 E i ), i 1 denes a pariion (E i ) i 1 of Ω s. Replacing E i by ( E i {ˆω i } ) {ˆω j : j 1, j = i} if necessary, we may assume ha ˆω i E i for i 1. We se ν i := ν (ˆωi) and P i := P (, ω s ˆω i, ν i ). Nex, we pase he conrols ν i. Fix N N and le A N := E 1 E N, hen {A c N, E1,..., E N } is a nie pariion of Ω s. Le X := X(s, ω, ν), dene ν(ω) := 1 [s,) ν(ω) + 1 [,T ] [ν(ω)1 A c N (X(ω) s ) + N i=1 ] ν i (ω )1 E i(x(ω) s ) and le P := P (s, ω, ν). Then, by Lemma 3.5, we have P = P on F s and P,ω = P (, ω s ω, ν i ) for all ω E i, for some subse E i E i of full measure P. Le us assume for he momen ha ˆω i E i for 1 i N, (3.13) 14

15 hen we may conclude ha P,ˆωi = P i for 1 i N. (3.14) Recall from Proposiion 2.5 ha V admis a modulus of coninuiy ρ (V). Moreover, we obain similarly as in (2.10) ha here exiss a modulus of coninuiy ρ (ξ) such ha ξ, ω sω ξ, ω sω ρ (ξ) ( ω ω [s,] ). Le ω E i Ω s for some 1 i N, hen ω ˆω i [s,] < ε. Togeher wih (3.12) and (3.14), we obain ha V s, ω (ω) V s, ω (ˆω i ) + ρ (V) (ε) Recall from (2.12) ha he mapping E P i [ξ, ω s ˆωi ] + ε + ρ (V) (ε) = E P,ˆωi [ξ, ω s ˆωi ] + ε + ρ (V) (ε). (3.15) ω E P (,ω,ν i) [ξ,ω ] is uniformly coninuous wih a modulus ρ independen of i and N. Since ω E i, i follows ha E P,ˆωi [ξ, ω s ˆωi ] E P,ω [ξ, ω sω ] = E P (, ω s ˆωi,ν i) [ξ, ω s ˆωi ] E P (, ω sω,νi) [ξ, ω sω ] ρ(ε) for P -a.e. ω E i. (3.16) Seing ρ(ε) := ρ(ε) + ε + ρ (V) (ε) and noing ha E P,ω [ξ, ω sω ] = E P,ω [(ξ s, ω ),ω ] = E P [ξ s, ω F s ](ω), he inequaliies (3.15) and (3.16) imply ha V s, ω (ω) E P [ ξ s, ω F s ] (ω) + ρ(ε) (3.17) for P -a.e. (and hus P -a.e.) ω E i. This holds for all 1 i N. As P = P on F s, aking P -expecaions yields E P [V s, ω 1 AN ] E P N [ξ s, ω 1 AN ] + ρ(ε), (3.18) where we wrie P N = P o recall he dependence on N. Since A N Ω s, we have P N (A c N ) = P (Ac N ) 0 as N. In view of E P N [ξ s, ω 1 AN ] = E P N [ξ s, ω ] E P N [ξ s, ω 1 A c N ] E P N [ξ s, ω ] + ξ P N (A c N), 15

16 we conclude from (3.18) ha E P [V s, ω ] lim sup E P N [ξ s, ω ] + ρ(ε) N sup E P [ξ s, ω ] + ρ(ε). P P(s, ω) Since P P(s, ω) was arbirary, leing ε 0 complees he proof of (3.11). I remains o argue ha our assumpion (3.13) does no enail a loss of generaliy. Indeed, assume ha ˆω i / E i for some i. Then here are wo possible cases. The case P (E i ) = 0 is easily seen o be harmless; recall ha he measure P was xed hroughou he proof. In he case P (E i ) > 0, we also have P ( E i ) > 0 and in paricular E i =. Thus we can replace ˆω i by an arbirary elemen of Ei (which can be chosen independenly of N). Using he coninuiy of he value funcion (Proposiion 2.5) and of he reward funcion (2.12), we see ha he above argumens sill apply if we add an addiional modulus of coninuiy in (3.15). 4 Exension of he Value Funcion In his secion, we exend he value funcion ξ V (ξ; ) o an L 1 -ype space of random variables ξ, in he spiri of, e.g., [8]. While he consrucion of V in he previous secion required a precise analysis ω by ω, we can now move owards a more probabilisic presenaion. In paricular, we shall ofen wrie V (ξ) for he random variable ω V (ξ; ω). For reasons explained in Remark 4.2 below, we x from now on an iniial condiion x R d and le P x := {P (0, x, ν) : ν U} be he corresponding se of measures a ime s = 0. Given a random variable ψ on Ω, we wrie ψ x as a shorhand for ψ 0,x ψ(x 0 ). We also wrie V x (ξ) for (V (ξ)) x. Given p [1, ), we dene L p P x o be he space of F T -measurable random variables X saisfying X L p Px := sup P P x X L p (P ) <, where X p L p (P ) := EP [ X p ]. More precisely, we idenify funcions which are equal P x -quasi-surely, so ha L p P x becomes a Banach space. (Two funcions are equal P x -quasi-surely, P x -q.s. for shor, if hey are equal P -a.s. for all P P x.) Furhermore, given [0, T ], L p P x (F ) is dened as he L p -closure of UC b (Ω ) L p P Px x. Since any L p P x -convergen sequence has a P x -q.s. convergen subsequence, any elemen of L p P x (F ) has an F -measurable represenaive. For breviy, we shall ofen wrie L p P x for L p P x (F T ). 16

17 Remark 4.1. The space L p P x can be described as follows. We say ha ξ L p P x is P x -quasi uniformly coninuous if ξ has a represenaive ξ wih he propery ha for all ε > 0 here exiss an open se G Ω such ha P (G) < ε for all P P and such ha he resricion ξ Ω G is uniformly coninuous. Then L p P x consiss of all ξ L p P x such ha ξ is P x -quasi uniformly coninuous and lim n ξ1 { ξ n} L p = 0. Moreover, If P x is weakly relaively compac, hen L p P x Px conains all bounded coninuous funcions on Ω. The proof is he same as in [17, Proposiion 5.2], which, in urn, followed an argumen of [7]. Before exending he value funcion o L 1 P x, le us explain why we are working under a xed iniial condiion x R d. Remark 4.2. There is no fundamenal obsrucion o wriing he heory wihou xing he iniial condiion x; in fac, mos of he resuls would be more elegan if saed using P insead of P x, where P is he se of all disribuions of he form (2.4), wih arbirary iniial condiion. However, he se P is very large and herefore he corresponding space L 1 P is very small, which is undesirable for he domain of our exended value funcion. As an illusraion, consider a random variable of he form ξ(ω) := f(ω 0 ) on Ω, where f : R d R is a measurable funcion. Then ξ L 1 P = sup x R d f(x) ; i.e., ξ is in L 1 P only when f is uniformly bounded. As a second issue in he same vein, i follows from he Arzelà-Ascoli heorem ha he se P is never weakly relaively compac. The laer propery, which is saised by P x for example when μ and σ are bounded, is someimes useful in he conex of quasi-sure analysis. Lemma 4.3. Le p [1, ). The mapping V x V x (ξ) V x (ψ) L p Px As a consequence, V x ξ ψ L p Px on UC b (Ω) is 1-Lipschiz, for all ξ, ψ UC b (Ω). uniquely exends o a Lipschiz-coninuous mapping V x : L p P x (F T ) L p P x (F ). Proof. The argumen is sandard and included only for compleeness. Noe ha ξ ψ p is again in UC b (Ω). The deniion of V x and Jensen's inequaliy imply ha V x (ξ) V x (ψ) p V x ( ξ ψ ) p V x ( ξ ψ p ). Therefore, V x (ξ) V x (ψ) L p Px sup P P x E P [ V x ( ξ ψ p ) ] 1/p = sup P P x E P [ ξ ψ p ] 1/p, where he equaliy is due o (3.2) applied wih s = 0. (For he case s = 0, he addiional Assumpion 3.1 was no used in he previous secion.) Recalling from Proposiion 2.5 ha V x maps UC b (Ω) o UC b (Ω ), i follows ha he exension maps L p P x o L p P x (F ). 17

18 4.1 Quasi-Sure Properies of he Exension In his secion, we provide some auxiliary resuls of echnical naure. The rs one will (quasi-surely) allow us o appeal o he resuls in he previous secion wihou imposing Assumpion 3.1. This is desirable since we would like o end up wih quasi-sure heorems whose saemens do no involve regular condiional probabiliy disribuions. Lemma 4.4. Assumpion 2.1 implies ha Assumpion 3.1 holds for P x - quasi-every η Ω saisfying η 0 = x. For he proof of his lemma, we shall use he following resul. Lemma 4.5. Le Y and Z be coninuous adaped processes, [0, T ] and le P be a probabiliy measure on Ω. Then F Y P F Z implies ha F Y,ω P,ω F Z,ω for P -a.e. ω Ω. Proof. The assumpion implies ha here exiss a progressively measurable ransformaion β : Ω Ω such ha Z = β(y ) P -a.s. For P -a.e. ω Ω, i follows ha Z(ω ) = β(y (ω )) P,ω -a.s., which, in urn, yields he resul. Proof of Lemma 4.4. Le X := X(, η, ν) wih η 0 = x, we have o show ha P 0 FX F whenever η 0 is ouside some P x -polar se. Hence we shall x an arbirary ˆP P x and show ha he resul holds on a se of full measure ˆP. Le ˆP P x, hen ˆP = P (0, x, ˆν) for some ˆν U and ˆP is concenraed on he image of ˆX 0, where ˆX := X(0, x, ˆν). Tha is, recalling ha η 0 = ˆX 0 = x, we may assume ha η = ˆX(ω) for some ω Ω. Le ν( ω) := 1 [0,)ˆν( ω) + 1 [,T ] ν( ω ). (4.1) Then X := X(0, x, ν) saises X = ˆX on [0, ] and hence we may assume ha η = X(ω) on [0, ]. Using Lemma 3.3 and (4.1), we deduce ha X,ω = X (, x X(ω), ν,ω) = X(, η, ν) = X. Since F F X P 0 by Assumpion 2.1, we conclude ha F P 0 F X,ω P 0 = F X P 0 by using Lemma 4.5 wih Z being he canonical process. The nex wo resuls show ha (for μ 0 and σ posiive denie) he mapping ξ V x (ξ) on L 1 P x falls ino he general class of sublinear expecaions considered in [19], whose echniques we shall apply in he subsequen 18

19 secion. More precisely, he wo lemmas below yield he validiy of is main condiion [19, Assumpion 4.1]. The following propery is known as sabiliy under pasing and well known o be imporan in non-markovian conrol. I should no be confused wih he pasing discussed in Remark 3.6, where he considered measures correspond o dieren poins in ime. Lemma 4.6. Le τ be an F-sopping ime and le Λ F τ. Le P, P 1, P 2 P x saisfy P 1 = P 2 = P on F τ. Then P (A) := E P [ P 1 (A F τ )1 Λ + P 2 ] (A F τ )1 Λ c, A FT denes an elemen of P x. Proof. I follows from he deniion of he condiional expecaion ha P is a probabiliy measure which is characerized by he properies { P = P on F τ and P τ(ω),ω (P 1 ) τ(ω),ω for P -a.e. ω Λ, = (P 2 ) τ(ω),ω for P -a.e. ω Λ c (4.2). Le ν, ν 1, ν 2 U be such ha P = P (0, x, ν) and P i = P (0, x, ν i ) for i = 1, 2. Moreover, le X := X(0, x, ν), dene ν U by ν r (ω) := 1 [0,τ(X(ω) 0 ))(r)ν r (ω) [ ] + 1 [τ(x(ω) 0 ),T ](r) νr 1 (ω)1 Λ (X(ω) 0 ) + νr 2 (ω)1 Λ c(x(ω) 0 ) and le P := P (0, x, ν) P x. We show ha P saises he hree properies from (4.2). Indeed, ν = ν on [0, τ(x 0 )) implies ha P = P on F τ. Moreover, as in (3.8), P τ(ω),ω = P ( τ(ω), x 0 ω, ν τ(x(ω)0 ),ˇω ) for P -a.e. ω Ω. Similarly as below (3.9), we also have ha ν τ(x(ω)0 ),ˇω = ν 1(ˇω τ(x(ω) 0 ) ) = (ν 1 ) τ(x(ω)0 ),ˇω for P -a.e. ω Λ. Therefore, P τ(ω),ω = P ( τ(ω), x 0 ω, (ν 1 ) τ(x(ω)0 ),ˇω ) = (P 1 ) τ(ω),ω for P -a.e. ω Λ. An analogous argumen esablishes he hird propery from (4.2) and we conclude ha P = P P x. The second propery is he quasi-sure represenaion of V x (ξ) on L 1 P x, a resul which will be generalized in Theorem 5.2 below. 19

20 Lemma 4.7. Le [0, T ] and ξ L 1 P x. Then V x (ξ) = ess sup P E P [ξ x F ] P -a.s. for all P P x, (4.3) P P x(f,p ) where P x (F, P ) := {P P x : P = P on F }. Proof. Recall ha Lemma 4.4 allows us o appeal o he resuls of Secion 3. (i) We rs prove he inequaliy for ξ UC b (Ω). Fix P P x. We use Sep (ii) of he proof of Theorem 3.2, in paricular (3.17), for he special case s = 0 and obain ha for given ε > 0 and N 1 here exiss a measure P N P x (F, P ) such ha V x (ω) E P N [ξ x F ](ω) + ρ(ε) for P -a.s. ω E 1 E N, where V x (ω) := V x (ξ; ω). Since i 1 Ei = Ω 0 = Ω P -a.s., we deduce ha V x (ω) sup E P N [ξ x F ](ω) + ρ(ε) for P -a.s. ω Ω. N 1 The claim follows by leing ε 0. (ii) Nex, we show he inequaliy in (4.3) for ξ UC b (Ω). Fix P, P P x and recall ha (P ),ω P(, x 0 ω) for P -a.s. ω Ω by Lemma 3.4. Therefore, (3.2) applied wih s := and := T yields ha V x (ω) = V (x 0 ω) E (P ),ω [ξ,x 0ω ] = E (P ),ω [(ξ x ),ω ] = E P [ξ x F s ](ω) P -a.s. on F. If P P x (F, P ), hen P = P on F and he inequaliy holds also P -a.s. The claim follows as P P x (F, P ) was arbirary. (iii) So far, we have proved he resul for ξ UC b (Ω). The general case ξ L 1 P x can be derived by an approximaion argumen exploiing he sabiliy under pasing (Lemma 4.6). We omi he deails since he proof is exacly he same as in [17, Theorem 5.4]. 5 Pah Regulariy for he Value Process In his secion, we consruc a càdlàg P x -modicaion for V x (ξ); ha is, a càdlàg process Y x such ha Y x = V x (ξ) P x -q.s. for all [0, T ]. (Recall ha he iniial condiion x R d has been xed.) To his end, we exend he raw lraion F as in [19]: we le F + = {F + } 0 T be he minimal righconinuous lraion conaining F and we augmen F + by he collecion N Px of (P x, F T )-polar ses o obain he lraion G = {G } 0 T, G := F + N Px. We noe ha G depends on x R d since N Px does, bu for breviy, we shall no indicae his in he noaion. In fac, he dependence on x is no crucial: 20

21 we could also work wih F +, a he expense of obaining a modicaion which is P x -q.s. equal o a càdlàg process raher han being càdlàg iself. We recall ha in he quasi-sure seing, value processes similar o he one under consideraion do no admi càdlàg modicaions in general; indeed, while he righ limi exiss quasi-surely, i need no be a modicaion (cf. [19]). Boh he regulariy of ξ L 1 P x and he regulariy induced by he SDE are crucial for he following resul. Theorem 5.1. Le ξ L 1 P x. There exiss a (P x -q.s. unique) G-adaped càdlàg P x -modicaion E x (ξ) = {E x (ξ)} [0,T ] of {V x (ξ)} [0,T ]. Moreover, E x (ξ) = for all [0, T ]. ess sup P E P [ξ x G ] P -a.s. for all P P x, (5.1) P P x(g,p ) Proof. In view of Lemmaa 4.6 and 4.7, we obain exacly as in [19, Proposiion 4.5] ha here exiss a P x -q.s. unique G-adaped càdlàg process E x (ξ) saisfying (5.1) and E x (ξ) = V+(ξ) x := lim Vr x (ξ) P x -q.s. for all 0 < T. (5.2) r The observaion made here is ha (4.3) implies ha V x (ξ) is a (P, F)- supermaringale for all P P x, so ha one can use he sandard modi- caion argumen for supermaringales under each P. This argumen, cf. [6, Theorem VI.2], also yields ha E P [E x (ξ) F + ] V x (ξ) P -a.s. and in paricular E P [E x (ξ)] E P [V x (ξ)] for all P P x. Hence, i remains o show ha E x (ξ) V x (ξ) P x -q.s. (5.3) for [0, T ), which is he par ha is known o fail in a more general seing. We give he proof in several seps. (i) We rs show ha E x maps UC b (Ω) o L 1 P x (F ), and in fac even o UC b (Ω ) if a suiable represenaive is chosen. Le ξ UC b (Ω), r (, T ] and se Vr x := Vr x (ξ). By Proposiion 2.5, here exiss a modulus of coninuiy ρ independen of r such ha V x r (ω) V x r (ω ) ρ( ω ω r ). Hence he P x -q.s. limi from (5.2) saises V + (ω) V + (ω ) ρ( ω ω r ) for all r (, T ] Q, P x -q.s. Since E x (ξ) = V + (ξ), aking he limi r yields ha E x (ξ)(ω) E x (ξ)(ω ) ρ( ω ω ) P x -q.s. 21

22 By a varian of Tieze's exension heorem, cf. [16], his implies ha E x (ξ) coincides P x -q.s. wih an elemen of UC b (Ω ). In paricular, E x (ξ) L 1 P x (F ). (ii) Nex, we show ha E x is Lipschiz-coninuous. Le ξ, ψ L 1 P x and n. Using (5.2), Faou's lemma and Lemma 4.3, we obain ha E x (ξ) E x (ψ) L 1 Px = limn V x n (ξ) V x n (ψ) L 1 Px = sup E P [ lim n V x n (ξ) V x n (ψ) ] P P x sup lim inf E P [ V x P P x n n (ξ) V x n (ψ) ] ξ ψ L 1 Px. (iii) Le ξ L 1 P x. Then here exis ξ n UC b (Ω) such ha ξ n ξ in L 1 P x and in hus E x (ξ n ) E x (ξ) in L 1 P x by Sep (ii). Since E x (ξ n ) L 1 P x (F ) by Sep (i) and since L 1 P x (F ) is closed in L 1 P x, we conclude ha E x (ξ) L 1 P x (F ) for all ξ L 1 P x. (iv) Le ξ L 1 P x. Since V x is he ideniy on L 1 P x (F ), Sep (iii) implies ha E x (ξ) = V x (E x (ξ)). Moreover, he represenaions (4.3) and (5.1) yield V x (E x (ξ)) = We conclude ha (5.3) holds rue. ess sup P E P [E x (ξ) F ] P P x(f,p ) ess sup P E P [ E P [ξ x G ] ] F P P x(f,p ) = ess sup P E P [ξ x F ] P P x(f,p ) = V x (ξ) P -a.s. for all P P x. Since E x (ξ) is a càdlàg process, is value E x τ (ξ) a a sopping ime τ is well dened. The following resul saes he quasi-sure represenaion of E x τ (ξ) and he quasi-sure version of he dynamic programming principle in is nal form. Theorem 5.2. Le 0 ρ τ T be G-sopping imes and ξ L 1 P x. Then E x ρ (ξ) = and in paricular E x ρ (ξ) = ess sup P E P [Eτ x (ξ) G ρ ] P -a.s. for all P P x (5.4) P P x(g ρ,p ) ess sup P E P [ξ x G ρ ] P -a.s. for all P P x. P P x(g ρ,p ) Moreover, here exiss for each P P x a sequence P n P x (G ρ, P ) such ha wih a P -a.s. increasing limi. E x ρ (ξ) = lim n EPn [ξ x G ρ ] 22 P -a.s.

23 Proof. In view of Lemmaa 4.6 and 4.7, he resul is derived exacly as in [19, Theorem 4.9]. As in (3.3), he relaion (5.4) can be seen as a semigroup propery E x ρ (ξ) = E x ρ (E x τ (ξ)), a leas when E x τ (ξ) is in he domain L 1 P x of E x ρ. The laer is guaraneed by Lemma 4.3 when τ is a deerminisic ime. However, one canno expec E x τ (ξ) o be quasi uniformly coninuous (cf. Remark 4.1) for a general sopping ime, for which reason we prefer o express he righ hand side as in (5.4). 6 Hamilon-Jacobi-Bellman 2BSDE In his secion, we characerize he value process E x (ξ) as he soluion of a 2BSDE. To his end, we rs examine he properies of B under a xed P P x. The following resul is in he spiri of [28, Secion 8]. Proposiion 6.1. Le x R d, ν U and P := P (0, x, ν). There exiss a progressively measurable ransformaion β : Ω Ω (depending on x, ν) such ha W := β(b) is a P -Brownian moion and F P = F W P. (6.1) Moreover, B is he P -a.s. unique srong soluion of he SDE B = 0 μ(, x + B, ν (W )) d + 0 σ(, x + B, ν (W )) dw under P. Proof. Le X := X(0, x, ν). As in Lemma 3.4, Assumpion 2.1 implies he he exisence of a progressively measurable ransformaion β : Ω Ω such ha β(x 0 ) = B P 0 -a.s. (6.2) Le W := β(b). Then (B, X 0 ) P0 = (β(x 0 ), X 0 ) P0 = (β(b), B) P = (W, B) P ; i.e., he disribuion of (B, X 0 ) under P 0 coincides wih he disribuion of (W, B) under P. In paricular, W is a P -Brownian moion. Moreover, we have F P X0 0 = F β(x0 ) P 0 by Assumpion 2.1 and herefore F BP = F β(b)p, which is (6.1). Noe ha X 0 = X 0 (B) = μ(, x + X 0, ν (B)) d + σ(, x + X 0, ν (B)) db 23

24 under P 0. Le Y be he (unique, srong) soluion of he analogous SDE Y = μ(, x + Y, ν (W )) d + σ(, x + Y, ν (W )) dw under P. Using he deniion of P and (6.2), we have ha (Y, W ) P = (X 0 (W ), W ) P = (X 0 (B), B) P0 = (X 0, β(x 0 )) P0 = (B, β(b)) P = (B, W ) P. In view of (6.1), i follows ha Y = B holds P -a.s. In he sequel, we denoe by M B,P he local maringale par in he canonical semimaringale decomposiion of B under P. Corollary 6.2. Le P P x. Then he lraion F P is righ-coninuous. If, in addiion, σ is inverible, hen (M B,P, P ) has he predicable represenaion propery. The laer saemen means ha any righ-coninuous (F P, P )-local maringale N has a represenaion N = N 0 + Z dm B,P under P, for some F P -predicable process Z. Proof. We have seen in Proposiion 6.1 ha F P is generaed by a Brownian moion W, hence righ-coninuous, and ha M B,P = 0 ˆσ dw for ˆσ := σ(, x + B, ν (W )), where ν U. By changing ˆσ on a d P -nullse, we may assume ha ˆσ is F P -predicable. Using he Brownian represenaion heorem and W = ˆσ 1 dm B,P, we deduce ha M B,P has he represenaion propery. The following formulaion of 2BSDE is, of course, inspired by [27]. Deniion 6.3. Le ξ L 1 P x and consider a pair (Y, Z) of processes wih values in R R d such ha Y is càdlàg G-adaped while Z is G-predicable and T 0 Z s 2 d B s < P x -q.s. Then (Y, Z) is called a soluion of he 2BSDE (6.3) if here exiss a family (K P ) P Px of F P -adaped increasing processes saisfying E P [ KT P ] < such ha Y = ξ T Z s dm B,P s + K P T K P, 0 T, P -a.s. for all P P x (6.3) and such ha he following minimaliy condiion holds for all 0 T : ess inf P [ P P x(g,p ) EP K P T K P ] G = 0 P -a.s. for all P Px. (6.4) 24

25 Moreover, a càdlàg process Y is said o be of class (D,P x ) if he family {Y τ } τ is uniformly inegrable under P for all P P x, where τ runs hrough all G-sopping imes. The following is our main resul. Theorem 6.4. Assume ha σ is inverible and le ξ L 1 P x. (i) There exiss a (d P x -q.s. unique) G-predicable process Z ξ such ha Z ξ = ( d B, B P ) 1 d E x (ξ), B P P -a.s. for all P P x. (6.5) (ii) The pair (E x (ξ), Z ξ ) is he minimal soluion of he 2BSDE (6.3); i.e., if (Y, Z) is anoher soluion, hen E x (ξ) Y P x -q.s. (iii) If (Y, Z) is a soluion of (6.3) such ha Y is of class (D,P x ), hen (Y, Z) = (E x (ξ), Z ξ ). In paricular, if ξ L p P x for some p > 1, hen (E x (ξ), Z ξ ) is he unique soluion of (6.3) in he class (D,P x ). Proof. Given wo processes which are (càdlàg) semimaringales under all P P x, heir quadraic covariaion can be dened P x -q.s. by using he inegraion-by-pars formula and Bicheler's pahwise sochasic inegraion [1, Theorem 7.14]; herefore, he righ hand side of (6.5) can be used as a deniion of Z ξ. The deails of he argumen are as in [19, Proposiion 4.10]. Le P P x. By Proposiion 6.1, B is an Iô process under P ; in paricular, we have B, S P = M B,P, S P P -a.s. for any P -semimaringale S. The Doob-Meyer heorem under P and Corollary 6.2 hen yield he decomposiion E x (ξ) = E0 x (ξ) + Z ξ dm B,P K P P -a.s. and we obain (ii) and (iii) by following he argumens in [19, Theorem 4.15]. If ξ L p P x for some p (1, ), hen E x (ξ) is of class (D,P x ) as a consequence of Jensen's inequaliy (cf. [19, Lemma 4.14]). Therefore, he las asserion follows from he above. We conclude by inerpreing he canonical process B, seen under he se of scenarios P x, as a model for drif and volailiy uncerainy in he Knighian sense. Remark 6.5. Consider he se-valued process D (ω) := {( μ(, ω, u), σ(, ω, u) ) : u U } R d R d d. In view of Proposiion 6.1, each P P x can be seen as a scenario in which he drif and he volailiy (of B) ake values in D, P -a.s. Then, he upper expecaion E x (ξ) is he corresponding wors-case expecaion (see [19] for a connecion o superhedging in nance). Noe ha D is a random process alhough he coeciens of our conrolled SDE are non-random. Indeed, 25

26 he pah-dependence of he SDE ranslaes o an ω-dependence in he weak formulaion ha we are considering. In paricular, for μ 0, we have consruced a sublinear expecaion similar o he random G-expecaion of [17]. While he laer is dened by specifying a se-valued process like D in he rs place, we have sared here from a conrolled SDE under P 0. I seems ha he presen consrucion is somewha less echnical ha he one in [17]; in paricular, we did no work wih he process ˆa = d B /d which played an imporan role in [26] and [17]. However, i seems ha he Lipschiz condiions on μ and σ are essenial, while [17] merely used a noion of uniform coninuiy. References [1] K. Bicheler. Sochasic inegraion and L p -heory of semimaringales. Ann. Probab., 9(1):4989, [2] R. Buckdahn and J. Ma. Sochasic viscosiy soluions for nonlinear sochasic parial dierenial equaions. I. Sochasic Process. Appl., 93(2):181204, [3] R. Buckdahn and J. Ma. Sochasic viscosiy soluions for nonlinear sochasic parial dierenial equaions. II. Sochasic Process. Appl., 93(2):205228, [4] R. Buckdahn and J. Ma. Pahwise sochasic conrol problems and sochasic HJB equaions. SIAM J. Conrol Opim., 45(6): , [5] P. Cheridio, H. M. Soner, N. Touzi, and N. Vicoir. Second-order backward sochasic dierenial equaions and fully nonlinear parabolic PDEs. Comm. Pure Appl. Mah., 60(7): , [6] C. Dellacherie and P. A. Meyer. Probabiliies and Poenial B. Norh Holland, Amserdam, [7] L. Denis, M. Hu, and S. Peng. Funcion spaces and capaciy relaed o a sublinear expecaion: applicaion o G-Brownian moion pahs. Poenial Anal., 34(2):139161, [8] L. Denis and C. Marini. A heoreical framework for he pricing of coningen claims in he presence of model uncerainy. Ann. Appl. Probab., 16(2): , [9] N. El Karoui. Les aspecs probabilises du conrôle sochasique. In Ecole d'éé de probabiliées de Sain-Flour, volume 876 of Lecure Noes in Mah., pages 73238, Springer, Berlin, [10] R. J. Ellio. Sochasic Calculus and Applicaions. Springer, New York, [11] J. Feldman and M. Smorodinsky. Simple examples of non-generaing Girsanov processes. In Séminaire de Probabiliés XXXI, volume 1655 of Lecure Noes in Mah., pages Springer, Berlin, [12] W. H. Fleming and H. M. Soner. Conrolled Markov Processes and Viscosiy Soluions. Springer, New York, 2nd ediion, [13] Y. Hu, P. Imkeller, and M. Müller. Uiliy maximizaion in incomplee markes. Ann. Appl. Probab., 15(3): , [14] P.-L. Lions and P. Souganidis. Fully nonlinear sochasic parial dierenial equaions. C. R. Acad. Sci. Paris Sér. I Mah., 326(9): , [15] P.-L. Lions and P. Souganidis. Fully nonlinear sochasic parial dierenial equaions: non-smooh equaions and applicaions. C. R. Acad. Sci. Paris Sér. I Mah., 327(8):735741,

27 [16] M. Mandelkern. On he uniform coninuiy of Tieze exensions. Arch. Mah., 55(4):387388, [17] M. Nuz. Random G-expecaions. Preprin arxiv: v1, [18] M. Nuz. The Bellman equaion for power uiliy maximizaion wih semimaringales. Ann. Appl. Probab., 22(1):363406, [19] M. Nuz and H. M. Soner. Superhedging and dynamic risk measures under volailiy uncerainy. Preprin arxiv: v1, [20] E. Pardoux and S. Peng. Adaped soluion of a backward sochasic dierenial equaion. Sysems Conrol Le., 14(1):5561, [21] S. Peng. Sochasic Hamilon-Jacobi-Bellman equaions. SIAM J. Conrol Opim., 30(2):284304, [22] S. Peng. Filraion consisen nonlinear expecaions and evaluaions of coningen claims. Aca Mah. Appl. Sin. Engl. Ser., 20(2):191214, [23] S. Peng. G-expecaion, G-Brownian moion and relaed sochasic calculus of Iô ype. In Sochasic Analysis and Applicaions, volume 2 of Abel Symp., pages , Springer, Berlin, [24] S. Peng. Muli-dimensional G-Brownian moion and relaed sochasic calculus under G-expecaion. Sochasic Process. Appl., 118(12): , [25] S. Peng. Noe on viscosiy soluion of pah-dependen PDE and G-maringales. Preprin arxiv: v1, [26] H. M. Soner, N. Touzi, and J. Zhang. Dual formulaion of second order arge problems. To appear in Ann. Appl. Probab., [27] H. M. Soner, N. Touzi, and J. Zhang. Wellposedness of second order backward SDEs. To appear in Probab. Theory Relaed Fields, [28] H. M. Soner, N. Touzi, and J. Zhang. Quasi-sure sochasic analysis hrough aggregaion. Elecron. J. Probab., 16(2): , [29] D. Sroock and S. R. S. Varadhan. Mulidimensional Diusion Processes. Springer, New York,

An Introduction to Backward Stochastic Differential Equations (BSDEs) PIMS Summer School 2016 in Mathematical Finance.

An Introduction to Backward Stochastic Differential Equations (BSDEs) PIMS Summer School 2016 in Mathematical Finance. 1 An Inroducion o Backward Sochasic Differenial Equaions (BSDEs) PIMS Summer School 2016 in Mahemaical Finance June 25, 2016 Chrisoph Frei cfrei@ualbera.ca This inroducion is based on Touzi [14], Bouchard

More information

Utility maximization in incomplete markets

Utility maximization in incomplete markets Uiliy maximizaion in incomplee markes Marcel Ladkau 27.1.29 Conens 1 Inroducion and general seings 2 1.1 Marke model....................................... 2 1.2 Trading sraegy.....................................

More information

Generalized Snell envelope and BSDE With Two general Reflecting Barriers

Generalized Snell envelope and BSDE With Two general Reflecting Barriers 1/22 Generalized Snell envelope and BSDE Wih Two general Reflecing Barriers EL HASSAN ESSAKY Cadi ayyad Universiy Poly-disciplinary Faculy Safi Work in progress wih : M. Hassani and Y. Ouknine Iasi, July

More information

arxiv: v1 [math.pr] 19 Feb 2011

arxiv: v1 [math.pr] 19 Feb 2011 A NOTE ON FELLER SEMIGROUPS AND RESOLVENTS VADIM KOSTRYKIN, JÜRGEN POTTHOFF, AND ROBERT SCHRADER ABSTRACT. Various equivalen condiions for a semigroup or a resolven generaed by a Markov process o be of

More information

An Introduction to Malliavin calculus and its applications

An Introduction to Malliavin calculus and its applications An Inroducion o Malliavin calculus and is applicaions Lecure 5: Smoohness of he densiy and Hörmander s heorem David Nualar Deparmen of Mahemaics Kansas Universiy Universiy of Wyoming Summer School 214

More information

Backward stochastic dynamics on a filtered probability space

Backward stochastic dynamics on a filtered probability space Backward sochasic dynamics on a filered probabiliy space Gechun Liang Oxford-Man Insiue, Universiy of Oxford based on join work wih Terry Lyons and Zhongmin Qian Page 1 of 15 gliang@oxford-man.ox.ac.uk

More information

Dual Representation as Stochastic Differential Games of Backward Stochastic Differential Equations and Dynamic Evaluations

Dual Representation as Stochastic Differential Games of Backward Stochastic Differential Equations and Dynamic Evaluations arxiv:mah/0602323v1 [mah.pr] 15 Feb 2006 Dual Represenaion as Sochasic Differenial Games of Backward Sochasic Differenial Equaions and Dynamic Evaluaions Shanjian Tang Absrac In his Noe, assuming ha he

More information

6. Stochastic calculus with jump processes

6. Stochastic calculus with jump processes A) Trading sraegies (1/3) Marke wih d asses S = (S 1,, S d ) A rading sraegy can be modelled wih a vecor φ describing he quaniies invesed in each asse a each insan : φ = (φ 1,, φ d ) The value a of a porfolio

More information

Quasi-sure Stochastic Analysis through Aggregation

Quasi-sure Stochastic Analysis through Aggregation E l e c r o n i c J o u r n a l o f P r o b a b i l i y Vol. 16 (211), Paper no. 67, pages 1844 1879. Journal URL hp://www.mah.washingon.edu/~ejpecp/ Quasi-sure Sochasic Analysis hrough Aggregaion H. Mee

More information

On a Fractional Stochastic Landau-Ginzburg Equation

On a Fractional Stochastic Landau-Ginzburg Equation Applied Mahemaical Sciences, Vol. 4, 1, no. 7, 317-35 On a Fracional Sochasic Landau-Ginzburg Equaion Nguyen Tien Dung Deparmen of Mahemaics, FPT Universiy 15B Pham Hung Sree, Hanoi, Vienam dungn@fp.edu.vn

More information

AMartingaleApproachforFractionalBrownian Motions and Related Path Dependent PDEs

AMartingaleApproachforFractionalBrownian Motions and Related Path Dependent PDEs AMaringaleApproachforFracionalBrownian Moions and Relaed Pah Dependen PDEs Jianfeng ZHANG Universiy of Souhern California Join work wih Frederi VIENS Mahemaical Finance, Probabiliy, and PDE Conference

More information

Uniqueness of solutions to quadratic BSDEs. BSDEs with convex generators and unbounded terminal conditions

Uniqueness of solutions to quadratic BSDEs. BSDEs with convex generators and unbounded terminal conditions Recalls and basic resuls on BSDEs Uniqueness resul Links wih PDEs On he uniqueness of soluions o quadraic BSDEs wih convex generaors and unbounded erminal condiions IRMAR, Universié Rennes 1 Châeau de

More information

A FAMILY OF MARTINGALES GENERATED BY A PROCESS WITH INDEPENDENT INCREMENTS

A FAMILY OF MARTINGALES GENERATED BY A PROCESS WITH INDEPENDENT INCREMENTS Theory of Sochasic Processes Vol. 14 3), no. 2, 28, pp. 139 144 UDC 519.21 JOSEP LLUÍS SOLÉ AND FREDERIC UTZET A FAMILY OF MARTINGALES GENERATED BY A PROCESS WITH INDEPENDENT INCREMENTS An explici procedure

More information

Lecture 10: The Poincaré Inequality in Euclidean space

Lecture 10: The Poincaré Inequality in Euclidean space Deparmens of Mahemaics Monana Sae Universiy Fall 215 Prof. Kevin Wildrick n inroducion o non-smooh analysis and geomery Lecure 1: The Poincaré Inequaliy in Euclidean space 1. Wha is he Poincaré inequaliy?

More information

Lecture Notes 2. The Hilbert Space Approach to Time Series

Lecture Notes 2. The Hilbert Space Approach to Time Series Time Series Seven N. Durlauf Universiy of Wisconsin. Basic ideas Lecure Noes. The Hilber Space Approach o Time Series The Hilber space framework provides a very powerful language for discussing he relaionship

More information

Monotonic Solutions of a Class of Quadratic Singular Integral Equations of Volterra type

Monotonic Solutions of a Class of Quadratic Singular Integral Equations of Volterra type In. J. Conemp. Mah. Sci., Vol. 2, 27, no. 2, 89-2 Monoonic Soluions of a Class of Quadraic Singular Inegral Equaions of Volerra ype Mahmoud M. El Borai Deparmen of Mahemaics, Faculy of Science, Alexandria

More information

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB Elecronic Companion EC.1. Proofs of Technical Lemmas and Theorems LEMMA 1. Le C(RB) be he oal cos incurred by he RB policy. Then we have, T L E[C(RB)] 3 E[Z RB ]. (EC.1) Proof of Lemma 1. Using he marginal

More information

4 Sequences of measurable functions

4 Sequences of measurable functions 4 Sequences of measurable funcions 1. Le (Ω, A, µ) be a measure space (complee, afer a possible applicaion of he compleion heorem). In his chaper we invesigae relaions beween various (nonequivalen) convergences

More information

A proof of Ito's formula using a di Title formula. Author(s) Fujita, Takahiko; Kawanishi, Yasuhi. Studia scientiarum mathematicarum H Citation

A proof of Ito's formula using a di Title formula. Author(s) Fujita, Takahiko; Kawanishi, Yasuhi. Studia scientiarum mathematicarum H Citation A proof of Io's formula using a di Tile formula Auhor(s) Fujia, Takahiko; Kawanishi, Yasuhi Sudia scieniarum mahemaicarum H Ciaion 15-134 Issue 8-3 Dae Type Journal Aricle Tex Version auhor URL hp://hdl.handle.ne/186/15878

More information

The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales

The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales Advances in Dynamical Sysems and Applicaions. ISSN 0973-5321 Volume 1 Number 1 (2006, pp. 103 112 c Research India Publicaions hp://www.ripublicaion.com/adsa.hm The Asympoic Behavior of Nonoscillaory Soluions

More information

EXISTENCE AND UNIQUENESS OF SOLUTIONS TO THE BACKWARD STOCHASTIC LORENZ SYSTEM

EXISTENCE AND UNIQUENESS OF SOLUTIONS TO THE BACKWARD STOCHASTIC LORENZ SYSTEM Communicaions on Sochasic Analysis Vol. 1, No. 3 (27) 473-483 EXISTENCE AND UNIQUENESS OF SOLUTIONS TO THE BACKWARD STOCHASTIC LORENZ SYSTEM P. SUNDAR AND HONG YIN Absrac. The backward sochasic Lorenz

More information

Homogenization of random Hamilton Jacobi Bellman Equations

Homogenization of random Hamilton Jacobi Bellman Equations Probabiliy, Geomery and Inegrable Sysems MSRI Publicaions Volume 55, 28 Homogenizaion of random Hamilon Jacobi Bellman Equaions S. R. SRINIVASA VARADHAN ABSTRACT. We consider nonlinear parabolic equaions

More information

1 Solutions to selected problems

1 Solutions to selected problems 1 Soluions o seleced problems 1. Le A B R n. Show ha in A in B bu in general bd A bd B. Soluion. Le x in A. Then here is ɛ > 0 such ha B ɛ (x) A B. This shows x in B. If A = [0, 1] and B = [0, 2], hen

More information

An introduction to the theory of SDDP algorithm

An introduction to the theory of SDDP algorithm An inroducion o he heory of SDDP algorihm V. Leclère (ENPC) Augus 1, 2014 V. Leclère Inroducion o SDDP Augus 1, 2014 1 / 21 Inroducion Large scale sochasic problem are hard o solve. Two ways of aacking

More information

On R d -valued peacocks

On R d -valued peacocks On R d -valued peacocks Francis HIRSCH 1), Bernard ROYNETTE 2) July 26, 211 1) Laboraoire d Analyse e Probabiliés, Universié d Évry - Val d Essonne, Boulevard F. Mierrand, F-9125 Évry Cedex e-mail: francis.hirsch@univ-evry.fr

More information

Couplage du principe des grandes déviations et de l homogénisation dans le cas des EDP paraboliques: (le cas constant)

Couplage du principe des grandes déviations et de l homogénisation dans le cas des EDP paraboliques: (le cas constant) Couplage du principe des grandes déviaions e de l homogénisaion dans le cas des EDP paraboliques: (le cas consan) Alioune COULIBALY U.F.R Sciences e Technologie Universié Assane SECK de Ziguinchor Probabilié

More information

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality Marix Versions of Some Refinemens of he Arihmeic-Geomeric Mean Inequaliy Bao Qi Feng and Andrew Tonge Absrac. We esablish marix versions of refinemens due o Alzer ], Carwrigh and Field 4], and Mercer 5]

More information

MODULE 3 FUNCTION OF A RANDOM VARIABLE AND ITS DISTRIBUTION LECTURES PROBABILITY DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE

MODULE 3 FUNCTION OF A RANDOM VARIABLE AND ITS DISTRIBUTION LECTURES PROBABILITY DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE Topics MODULE 3 FUNCTION OF A RANDOM VARIABLE AND ITS DISTRIBUTION LECTURES 2-6 3. FUNCTION OF A RANDOM VARIABLE 3.2 PROBABILITY DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE 3.3 EXPECTATION AND MOMENTS

More information

Hamilton- J acobi Equation: Weak S olution We continue the study of the Hamilton-Jacobi equation:

Hamilton- J acobi Equation: Weak S olution We continue the study of the Hamilton-Jacobi equation: M ah 5 7 Fall 9 L ecure O c. 4, 9 ) Hamilon- J acobi Equaion: Weak S oluion We coninue he sudy of he Hamilon-Jacobi equaion: We have shown ha u + H D u) = R n, ) ; u = g R n { = }. ). In general we canno

More information

Finish reading Chapter 2 of Spivak, rereading earlier sections as necessary. handout and fill in some missing details!

Finish reading Chapter 2 of Spivak, rereading earlier sections as necessary. handout and fill in some missing details! MAT 257, Handou 6: Ocober 7-2, 20. I. Assignmen. Finish reading Chaper 2 of Spiva, rereading earlier secions as necessary. handou and fill in some missing deails! II. Higher derivaives. Also, read his

More information

Cash Flow Valuation Mode Lin Discrete Time

Cash Flow Valuation Mode Lin Discrete Time IOSR Journal of Mahemaics (IOSR-JM) e-issn: 2278-5728,p-ISSN: 2319-765X, 6, Issue 6 (May. - Jun. 2013), PP 35-41 Cash Flow Valuaion Mode Lin Discree Time Olayiwola. M. A. and Oni, N. O. Deparmen of Mahemaics

More information

Existence of positive solution for a third-order three-point BVP with sign-changing Green s function

Existence of positive solution for a third-order three-point BVP with sign-changing Green s function Elecronic Journal of Qualiaive Theory of Differenial Equaions 13, No. 3, 1-11; hp://www.mah.u-szeged.hu/ejqde/ Exisence of posiive soluion for a hird-order hree-poin BVP wih sign-changing Green s funcion

More information

Mean-Variance Hedging for General Claims

Mean-Variance Hedging for General Claims Projekbereich B Discussion Paper No. B 167 Mean-Variance Hedging for General Claims by Marin Schweizer ) Ocober 199 ) Financial suppor by Deusche Forschungsgemeinschaf, Sonderforschungsbereich 33 a he

More information

Time discretization of quadratic and superquadratic Markovian BSDEs with unbounded terminal conditions

Time discretization of quadratic and superquadratic Markovian BSDEs with unbounded terminal conditions Time discreizaion of quadraic and superquadraic Markovian BSDEs wih unbounded erminal condiions Adrien Richou Universié Bordeaux 1, INRIA équipe ALEA Oxford framework Le (Ω, F, P) be a probabiliy space,

More information

REFLECTED SOLUTIONS OF BACKWARD SDE S, AND RELATED OBSTACLE PROBLEMS FOR PDE S

REFLECTED SOLUTIONS OF BACKWARD SDE S, AND RELATED OBSTACLE PROBLEMS FOR PDE S The Annals of Probabiliy 1997, Vol. 25, No. 2, 72 737 REFLECTED SOLUTIONS OF BACKWARD SDE S, AND RELATED OBSTACLE PROBLEMS FOR PDE S By N. El Karoui, C. Kapoudjian, E. Pardoux, S. Peng and M. C. Quenez

More information

BACKWARD STOCHASTIC DIFFERENTIAL EQUATIONS WITH REFLECTION AND DYNKIN GAMES 1. By Jakša Cvitanić and Ioannis Karatzas Columbia University

BACKWARD STOCHASTIC DIFFERENTIAL EQUATIONS WITH REFLECTION AND DYNKIN GAMES 1. By Jakša Cvitanić and Ioannis Karatzas Columbia University The Annals of Probabiliy 1996, Vol. 24, No. 4, 224 256 BACKWARD STOCHASTIC DIFFERENTIAL EQUATIONS WITH REFLECTION AND DYNKIN GAMES 1 By Jakša Cvianić and Ioannis Karazas Columbia Universiy We esablish

More information

Risk Aversion Asymptotics for Power Utility Maximization

Risk Aversion Asymptotics for Power Utility Maximization Risk Aversion Asympoics for Power Uiliy Maximizaion Marcel Nuz ETH Zurich AnSAp10 Conference Vienna, 12.07.2010 Marcel Nuz (ETH) Risk Aversion Asympoics 1 / 15 Basic Problem Power uiliy funcion U(x) =

More information

11!Hí MATHEMATICS : ERDŐS AND ULAM PROC. N. A. S. of decomposiion, properly speaking) conradics he possibiliy of defining a counably addiive real-valu

11!Hí MATHEMATICS : ERDŐS AND ULAM PROC. N. A. S. of decomposiion, properly speaking) conradics he possibiliy of defining a counably addiive real-valu ON EQUATIONS WITH SETS AS UNKNOWNS BY PAUL ERDŐS AND S. ULAM DEPARTMENT OF MATHEMATICS, UNIVERSITY OF COLORADO, BOULDER Communicaed May 27, 1968 We shall presen here a number of resuls in se heory concerning

More information

arxiv: v1 [math.fa] 9 Dec 2018

arxiv: v1 [math.fa] 9 Dec 2018 AN INVERSE FUNCTION THEOREM CONVERSE arxiv:1812.03561v1 [mah.fa] 9 Dec 2018 JIMMIE LAWSON Absrac. We esablish he following converse of he well-known inverse funcion heorem. Le g : U V and f : V U be inverse

More information

Semilinear Kolmogorov equations and applications to stochastic optimal control

Semilinear Kolmogorov equations and applications to stochastic optimal control Semilinear Kolmogorov equaions and applicaions o sochasic opimal conrol Federica Masiero 1 Advisor: Prof. Marco Fuhrman 2 1 Diparimeno di Maemaica, Universià degli sudi di Milano, Via Saldini 5, 2133 Milano,

More information

Backward doubly stochastic di erential equations with quadratic growth and applications to quasilinear SPDEs

Backward doubly stochastic di erential equations with quadratic growth and applications to quasilinear SPDEs Backward doubly sochasic di erenial equaions wih quadraic growh and applicaions o quasilinear SPDEs Badreddine MANSOURI (wih K. Bahlali & B. Mezerdi) Universiy of Biskra Algeria La Londe 14 sepember 2007

More information

Quadratic and Superquadratic BSDEs and Related PDEs

Quadratic and Superquadratic BSDEs and Related PDEs Quadraic and Superquadraic BSDEs and Relaed PDEs Ying Hu IRMAR, Universié Rennes 1, FRANCE hp://perso.univ-rennes1.fr/ying.hu/ ITN Marie Curie Workshop "Sochasic Conrol and Finance" Roscoff, March 21 Ying

More information

On Gronwall s Type Integral Inequalities with Singular Kernels

On Gronwall s Type Integral Inequalities with Singular Kernels Filoma 31:4 (217), 141 149 DOI 1.2298/FIL17441A Published by Faculy of Sciences and Mahemaics, Universiy of Niš, Serbia Available a: hp://www.pmf.ni.ac.rs/filoma On Gronwall s Type Inegral Inequaliies

More information

LECTURE 1: GENERALIZED RAY KNIGHT THEOREM FOR FINITE MARKOV CHAINS

LECTURE 1: GENERALIZED RAY KNIGHT THEOREM FOR FINITE MARKOV CHAINS LECTURE : GENERALIZED RAY KNIGHT THEOREM FOR FINITE MARKOV CHAINS We will work wih a coninuous ime reversible Markov chain X on a finie conneced sae space, wih generaor Lf(x = y q x,yf(y. (Recall ha q

More information

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes Represening Periodic Funcions by Fourier Series 3. Inroducion In his Secion we show how a periodic funcion can be expressed as a series of sines and cosines. We begin by obaining some sandard inegrals

More information

Markov Processes and Stochastic Calculus

Markov Processes and Stochastic Calculus Markov Processes and Sochasic Calculus René Caldeney In his noes we revise he basic noions of Brownian moions, coninuous ime Markov processes and sochasic differenial equaions in he Iô sense. 1 Inroducion

More information

Lecture 20: Riccati Equations and Least Squares Feedback Control

Lecture 20: Riccati Equations and Least Squares Feedback Control 34-5 LINEAR SYSTEMS Lecure : Riccai Equaions and Leas Squares Feedback Conrol 5.6.4 Sae Feedback via Riccai Equaions A recursive approach in generaing he marix-valued funcion W ( ) equaion for i for he

More information

STABILITY OF PEXIDERIZED QUADRATIC FUNCTIONAL EQUATION IN NON-ARCHIMEDEAN FUZZY NORMED SPASES

STABILITY OF PEXIDERIZED QUADRATIC FUNCTIONAL EQUATION IN NON-ARCHIMEDEAN FUZZY NORMED SPASES Novi Sad J. Mah. Vol. 46, No. 1, 2016, 15-25 STABILITY OF PEXIDERIZED QUADRATIC FUNCTIONAL EQUATION IN NON-ARCHIMEDEAN FUZZY NORMED SPASES N. Eghbali 1 Absrac. We deermine some sabiliy resuls concerning

More information

arxiv: v1 [math.pr] 18 Feb 2015

arxiv: v1 [math.pr] 18 Feb 2015 Non-Markovian opimal sopping problems and consrained BSDEs wih jump arxiv:152.5422v1 [mah.pr 18 Feb 215 Marco Fuhrman Poliecnico di Milano, Diparimeno di Maemaica via Bonardi 9, 2133 Milano, Ialy marco.fuhrman@polimi.i

More information

Chapter 2. First Order Scalar Equations

Chapter 2. First Order Scalar Equations Chaper. Firs Order Scalar Equaions We sar our sudy of differenial equaions in he same way he pioneers in his field did. We show paricular echniques o solve paricular ypes of firs order differenial equaions.

More information

SOME MORE APPLICATIONS OF THE HAHN-BANACH THEOREM

SOME MORE APPLICATIONS OF THE HAHN-BANACH THEOREM SOME MORE APPLICATIONS OF THE HAHN-BANACH THEOREM FRANCISCO JAVIER GARCÍA-PACHECO, DANIELE PUGLISI, AND GUSTI VAN ZYL Absrac We give a new proof of he fac ha equivalen norms on subspaces can be exended

More information

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems MATHEMATICS OF OPERATIONS RESEARCH Vol. 38, No. 2, May 2013, pp. 209 227 ISSN 0364-765X (prin) ISSN 1526-5471 (online) hp://dx.doi.org/10.1287/moor.1120.0562 2013 INFORMS On Boundedness of Q-Learning Ieraes

More information

Approximation of backward stochastic variational inequalities

Approximation of backward stochastic variational inequalities Al. I. Cuza Universiy of Iaşi, România 10ème Colloque Franco-Roumain de Mahémaiques Appliquées Augus 27, 2010, Poiiers, France Shor hisory & moivaion Re eced Sochasic Di erenial Equaions were rs sudied

More information

Convergence of the Neumann series in higher norms

Convergence of the Neumann series in higher norms Convergence of he Neumann series in higher norms Charles L. Epsein Deparmen of Mahemaics, Universiy of Pennsylvania Version 1.0 Augus 1, 003 Absrac Naural condiions on an operaor A are given so ha he Neumann

More information

Stability and Bifurcation in a Neural Network Model with Two Delays

Stability and Bifurcation in a Neural Network Model with Two Delays Inernaional Mahemaical Forum, Vol. 6, 11, no. 35, 175-1731 Sabiliy and Bifurcaion in a Neural Nework Model wih Two Delays GuangPing Hu and XiaoLing Li School of Mahemaics and Physics, Nanjing Universiy

More information

BSDES UNDER FILTRATION-CONSISTENT NONLINEAR EXPECTATIONS AND THE CORRESPONDING DECOMPOSITION THEOREM FOR E-SUPERMARTINGALES IN L p

BSDES UNDER FILTRATION-CONSISTENT NONLINEAR EXPECTATIONS AND THE CORRESPONDING DECOMPOSITION THEOREM FOR E-SUPERMARTINGALES IN L p ROCKY MOUNTAIN JOURNAL OF MATHEMATICS Volume 43, Number 2, 213 BSDES UNDER FILTRATION-CONSISTENT NONLINEAR EXPECTATIONS AND THE CORRESPONDING DECOMPOSITION THEOREM FOR E-SUPERMARTINGALES IN L p ZHAOJUN

More information

CONTRIBUTION TO IMPULSIVE EQUATIONS

CONTRIBUTION TO IMPULSIVE EQUATIONS European Scienific Journal Sepember 214 /SPECIAL/ ediion Vol.3 ISSN: 1857 7881 (Prin) e - ISSN 1857-7431 CONTRIBUTION TO IMPULSIVE EQUATIONS Berrabah Faima Zohra, MA Universiy of sidi bel abbes/ Algeria

More information

Vehicle Arrival Models : Headway

Vehicle Arrival Models : Headway Chaper 12 Vehicle Arrival Models : Headway 12.1 Inroducion Modelling arrival of vehicle a secion of road is an imporan sep in raffic flow modelling. I has imporan applicaion in raffic flow simulaion where

More information

Singular control of SPDEs and backward stochastic partial diffe. reflection

Singular control of SPDEs and backward stochastic partial diffe. reflection Singular conrol of SPDEs and backward sochasic parial differenial equaions wih reflecion Universiy of Mancheser Join work wih Bern Øksendal and Agnès Sulem Singular conrol of SPDEs and backward sochasic

More information

1 Inroducion T. Lyons and T.S. Zhang ([1]) considered a saionary good symmeric Markov process (X(); ) associaed wih a Dirichle form and showed ha, for

1 Inroducion T. Lyons and T.S. Zhang ([1]) considered a saionary good symmeric Markov process (X(); ) associaed wih a Dirichle form and showed ha, for A generalized class of Lyons-Zheng processes Francesco Russo (1) Pierre Vallois () Jochen Wolf (1) (1) Universie Paris 13 Insiu Galilee, Mahemaiques Avenue J.B. Clemen F-9343 Villeaneuse () Universie de

More information

arxiv: v1 [math.pr] 6 Oct 2008

arxiv: v1 [math.pr] 6 Oct 2008 MEASURIN THE NON-STOPPIN TIMENESS OF ENDS OF PREVISIBLE SETS arxiv:8.59v [mah.pr] 6 Oc 8 JU-YI YEN ),) AND MARC YOR 3),4) Absrac. In his paper, we propose several measuremens of he nonsopping imeness of

More information

Martingales Stopping Time Processes

Martingales Stopping Time Processes IOSR Journal of Mahemaics (IOSR-JM) e-issn: 2278-5728, p-issn: 2319-765. Volume 11, Issue 1 Ver. II (Jan - Feb. 2015), PP 59-64 www.iosrjournals.org Maringales Sopping Time Processes I. Fulaan Deparmen

More information

Hamilton- J acobi Equation: Explicit Formulas In this lecture we try to apply the method of characteristics to the Hamilton-Jacobi equation: u t

Hamilton- J acobi Equation: Explicit Formulas In this lecture we try to apply the method of characteristics to the Hamilton-Jacobi equation: u t M ah 5 2 7 Fall 2 0 0 9 L ecure 1 0 O c. 7, 2 0 0 9 Hamilon- J acobi Equaion: Explici Formulas In his lecure we ry o apply he mehod of characerisics o he Hamilon-Jacobi equaion: u + H D u, x = 0 in R n

More information

Ann. Funct. Anal. 2 (2011), no. 2, A nnals of F unctional A nalysis ISSN: (electronic) URL:

Ann. Funct. Anal. 2 (2011), no. 2, A nnals of F unctional A nalysis ISSN: (electronic) URL: Ann. Func. Anal. 2 2011, no. 2, 34 41 A nnals of F uncional A nalysis ISSN: 2008-8752 elecronic URL: www.emis.de/journals/afa/ CLASSIFICAION OF POSIIVE SOLUIONS OF NONLINEAR SYSEMS OF VOLERRA INEGRAL EQUAIONS

More information

Optimal Control versus Stochastic Target problems: An Equivalence Result

Optimal Control versus Stochastic Target problems: An Equivalence Result Opimal Conrol versus Sochasic Targe problems: An quivalence Resul Bruno Bouchard CRMAD, Universié Paris Dauphine and CRST-NSA bouchard@ceremade.dauphine.fr Ngoc Minh Dang CRMAD, Universié Paris Dauphine

More information

Local Strict Comparison Theorem and Converse Comparison Theorems for Reflected Backward Stochastic Differential Equations

Local Strict Comparison Theorem and Converse Comparison Theorems for Reflected Backward Stochastic Differential Equations arxiv:mah/07002v [mah.pr] 3 Dec 2006 Local Sric Comparison Theorem and Converse Comparison Theorems for Refleced Backward Sochasic Differenial Equaions Juan Li and Shanjian Tang Absrac A local sric comparison

More information

arxiv: v1 [math.pr] 28 Nov 2016

arxiv: v1 [math.pr] 28 Nov 2016 Backward Sochasic Differenial Equaions wih Nonmarkovian Singular Terminal Values Ali Devin Sezer, Thomas Kruse, Alexandre Popier Ocober 15, 2018 arxiv:1611.09022v1 mah.pr 28 Nov 2016 Absrac We solve a

More information

International Journal of Scientific & Engineering Research, Volume 4, Issue 10, October ISSN

International Journal of Scientific & Engineering Research, Volume 4, Issue 10, October ISSN Inernaional Journal of Scienific & Engineering Research, Volume 4, Issue 10, Ocober-2013 900 FUZZY MEAN RESIDUAL LIFE ORDERING OF FUZZY RANDOM VARIABLES J. EARNEST LAZARUS PIRIYAKUMAR 1, A. YAMUNA 2 1.

More information

Lecture 6: Wiener Process

Lecture 6: Wiener Process Lecure 6: Wiener Process Eric Vanden-Eijnden Chapers 6, 7 and 8 offer a (very) brief inroducion o sochasic analysis. These lecures are based in par on a book projec wih Weinan E. A sandard reference for

More information

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation Course Noes for EE7C Spring 018: Convex Opimizaion and Approximaion Insrucor: Moriz Hard Email: hard+ee7c@berkeley.edu Graduae Insrucor: Max Simchowiz Email: msimchow+ee7c@berkeley.edu Ocober 15, 018 3

More information

ODEs II, Lecture 1: Homogeneous Linear Systems - I. Mike Raugh 1. March 8, 2004

ODEs II, Lecture 1: Homogeneous Linear Systems - I. Mike Raugh 1. March 8, 2004 ODEs II, Lecure : Homogeneous Linear Sysems - I Mike Raugh March 8, 4 Inroducion. In he firs lecure we discussed a sysem of linear ODEs for modeling he excreion of lead from he human body, saw how o ransform

More information

A remark on the H -calculus

A remark on the H -calculus A remark on he H -calculus Nigel J. Kalon Absrac If A, B are secorial operaors on a Hilber space wih he same domain range, if Ax Bx A 1 x B 1 x, hen i is a resul of Auscher, McInosh Nahmod ha if A has

More information

IMPLICIT AND INVERSE FUNCTION THEOREMS PAUL SCHRIMPF 1 OCTOBER 25, 2013

IMPLICIT AND INVERSE FUNCTION THEOREMS PAUL SCHRIMPF 1 OCTOBER 25, 2013 IMPLICI AND INVERSE FUNCION HEOREMS PAUL SCHRIMPF 1 OCOBER 25, 213 UNIVERSIY OF BRIISH COLUMBIA ECONOMICS 526 We have exensively sudied how o solve sysems of linear equaions. We know how o check wheher

More information

Differential Harnack Estimates for Parabolic Equations

Differential Harnack Estimates for Parabolic Equations Differenial Harnack Esimaes for Parabolic Equaions Xiaodong Cao and Zhou Zhang Absrac Le M,g be a soluion o he Ricci flow on a closed Riemannian manifold In his paper, we prove differenial Harnack inequaliies

More information

Positive continuous solution of a quadratic integral equation of fractional orders

Positive continuous solution of a quadratic integral equation of fractional orders Mah. Sci. Le., No., 9-7 (3) 9 Mahemaical Sciences Leers An Inernaional Journal @ 3 NSP Naural Sciences Publishing Cor. Posiive coninuous soluion of a quadraic inegral equaion of fracional orders A. M.

More information

Existence Theory of Second Order Random Differential Equations

Existence Theory of Second Order Random Differential Equations Global Journal of Mahemaical Sciences: Theory and Pracical. ISSN 974-32 Volume 4, Number 3 (22), pp. 33-3 Inernaional Research Publicaion House hp://www.irphouse.com Exisence Theory of Second Order Random

More information

L p -L q -Time decay estimate for solution of the Cauchy problem for hyperbolic partial differential equations of linear thermoelasticity

L p -L q -Time decay estimate for solution of the Cauchy problem for hyperbolic partial differential equations of linear thermoelasticity ANNALES POLONICI MATHEMATICI LIV.2 99) L p -L q -Time decay esimae for soluion of he Cauchy problem for hyperbolic parial differenial equaions of linear hermoelasiciy by Jerzy Gawinecki Warszawa) Absrac.

More information

Essential Maps and Coincidence Principles for General Classes of Maps

Essential Maps and Coincidence Principles for General Classes of Maps Filoma 31:11 (2017), 3553 3558 hps://doi.org/10.2298/fil1711553o Published by Faculy of Sciences Mahemaics, Universiy of Niš, Serbia Available a: hp://www.pmf.ni.ac.rs/filoma Essenial Maps Coincidence

More information

MATH 4330/5330, Fourier Analysis Section 6, Proof of Fourier s Theorem for Pointwise Convergence

MATH 4330/5330, Fourier Analysis Section 6, Proof of Fourier s Theorem for Pointwise Convergence MATH 433/533, Fourier Analysis Secion 6, Proof of Fourier s Theorem for Poinwise Convergence Firs, some commens abou inegraing periodic funcions. If g is a periodic funcion, g(x + ) g(x) for all real x,

More information

f(s)dw Solution 1. Approximate f by piece-wise constant left-continuous non-random functions f n such that (f(s) f n (s)) 2 ds 0.

f(s)dw Solution 1. Approximate f by piece-wise constant left-continuous non-random functions f n such that (f(s) f n (s)) 2 ds 0. Advanced Financial Models Example shee 3 - Michaelmas 217 Michael Tehranchi Problem 1. Le f : [, R be a coninuous (non-random funcion and W a Brownian moion, and le σ 2 = f(s 2 ds and assume σ 2

More information

Weyl sequences: Asymptotic distributions of the partition lengths

Weyl sequences: Asymptotic distributions of the partition lengths ACTA ARITHMETICA LXXXVIII.4 (999 Weyl sequences: Asympoic disribuions of he pariion lenghs by Anaoly Zhigljavsky (Cardiff and Iskander Aliev (Warszawa. Inroducion: Saemen of he problem and formulaion of

More information

15. Vector Valued Functions

15. Vector Valued Functions 1. Vecor Valued Funcions Up o his poin, we have presened vecors wih consan componens, for example, 1, and,,4. However, we can allow he componens of a vecor o be funcions of a common variable. For example,

More information

Dedicated to the memory of Professor Dragoslav S. Mitrinovic 1. INTRODUCTION. Let E :[0;+1)!Rbe a nonnegative, non-increasing, locally absolutely

Dedicated to the memory of Professor Dragoslav S. Mitrinovic 1. INTRODUCTION. Let E :[0;+1)!Rbe a nonnegative, non-increasing, locally absolutely Univ. Beograd. Publ. Elekroehn. Fak. Ser. Ma. 7 (1996), 55{67. DIFFERENTIAL AND INTEGRAL INEQUALITIES Vilmos Komornik Dedicaed o he memory of Professor Dragoslav S. Mirinovic 1. INTRODUCTION Le E :[;)!Rbe

More information

Optima and Equilibria for Traffic Flow on a Network

Optima and Equilibria for Traffic Flow on a Network Opima and Equilibria for Traffic Flow on a Nework Albero Bressan Deparmen of Mahemaics, Penn Sae Universiy bressan@mah.psu.edu Albero Bressan (Penn Sae) Opima and equilibria for raffic flow 1 / 1 A Traffic

More information

A NOTE ON THE SMOLUCHOWSKI-KRAMERS APPROXIMATION FOR THE LANGEVIN EQUATION WITH REFLECTION

A NOTE ON THE SMOLUCHOWSKI-KRAMERS APPROXIMATION FOR THE LANGEVIN EQUATION WITH REFLECTION A NOTE ON THE SMOLUCHOWSKI-KRAMERS APPROXIMATION FOR THE LANGEVIN EQUATION WITH REFLECTION KONSTANTINOS SPILIOPOULOS Deparmen of Mahemaics, Universiy of Maryland, College Park, 2742, Maryland, USA kspiliop@mah.umd.edu

More information

Kalman Bucy filtering equations of forward and backward stochastic systems and applications to recursive optimal control problems

Kalman Bucy filtering equations of forward and backward stochastic systems and applications to recursive optimal control problems J. Mah. Anal. Appl. 34 8) 18 196 www.elsevier.com/locae/jmaa Kalman Bucy filering equaions of forward and backward sochasic sysems and applicaions o recursive opimal conrol problems Guangchen Wang a,b,,zhenwu

More information

Asymptotic instability of nonlinear differential equations

Asymptotic instability of nonlinear differential equations Elecronic Journal of Differenial Equaions, Vol. 1997(1997), No. 16, pp. 1 7. ISSN: 172-6691. URL: hp://ejde.mah.sw.edu or hp://ejde.mah.un.edu fp (login: fp) 147.26.13.11 or 129.12.3.113 Asympoic insabiliy

More information

Some New Uniqueness Results of Solutions to Nonlinear Fractional Integro-Differential Equations

Some New Uniqueness Results of Solutions to Nonlinear Fractional Integro-Differential Equations Annals of Pure and Applied Mahemaics Vol. 6, No. 2, 28, 345-352 ISSN: 2279-87X (P), 2279-888(online) Published on 22 February 28 www.researchmahsci.org DOI: hp://dx.doi.org/.22457/apam.v6n2a Annals of

More information

BY PAWE L HITCZENKO Department of Mathematics, Box 8205, North Carolina State University, Raleigh, NC , USA

BY PAWE L HITCZENKO Department of Mathematics, Box 8205, North Carolina State University, Raleigh, NC , USA Absrac Tangen Sequences in Orlicz and Rearrangemen Invarian Spaces BY PAWE L HITCZENKO Deparmen of Mahemaics, Box 8205, Norh Carolina Sae Universiy, Raleigh, NC 27695 8205, USA AND STEPHEN J MONTGOMERY-SMITH

More information

The Optimal Stopping Time for Selling an Asset When It Is Uncertain Whether the Price Process Is Increasing or Decreasing When the Horizon Is Infinite

The Optimal Stopping Time for Selling an Asset When It Is Uncertain Whether the Price Process Is Increasing or Decreasing When the Horizon Is Infinite American Journal of Operaions Research, 08, 8, 8-9 hp://wwwscirporg/journal/ajor ISSN Online: 60-8849 ISSN Prin: 60-8830 The Opimal Sopping Time for Selling an Asse When I Is Uncerain Wheher he Price Process

More information

A class of multidimensional quadratic BSDEs

A class of multidimensional quadratic BSDEs A class of mulidimensional quadraic SDEs Zhongmin Qian, Yimin Yang Shujin Wu March 4, 07 arxiv:703.0453v mah.p] Mar 07 Absrac In his paper we sudy a mulidimensional quadraic SDE wih a paricular class of

More information

Vanishing Viscosity Method. There are another instructive and perhaps more natural discontinuous solutions of the conservation law

Vanishing Viscosity Method. There are another instructive and perhaps more natural discontinuous solutions of the conservation law Vanishing Viscosiy Mehod. There are anoher insrucive and perhaps more naural disconinuous soluions of he conservaion law (1 u +(q(u x 0, he so called vanishing viscosiy mehod. This mehod consiss in viewing

More information

2. Nonlinear Conservation Law Equations

2. Nonlinear Conservation Law Equations . Nonlinear Conservaion Law Equaions One of he clear lessons learned over recen years in sudying nonlinear parial differenial equaions is ha i is generally no wise o ry o aack a general class of nonlinear

More information

Explicit construction of a dynamic Bessel bridge of dimension 3

Explicit construction of a dynamic Bessel bridge of dimension 3 Explici consrucion of a dynamic Bessel bridge of dimension 3 Luciano Campi Umu Çein Albina Danilova February 25, 23 Absrac Given a deerminisically ime-changed Brownian moion Z saring from, whose imechange

More information

Class Meeting # 10: Introduction to the Wave Equation

Class Meeting # 10: Introduction to the Wave Equation MATH 8.5 COURSE NOTES - CLASS MEETING # 0 8.5 Inroducion o PDEs, Fall 0 Professor: Jared Speck Class Meeing # 0: Inroducion o he Wave Equaion. Wha is he wave equaion? The sandard wave equaion for a funcion

More information

Heavy Tails of Discounted Aggregate Claims in the Continuous-time Renewal Model

Heavy Tails of Discounted Aggregate Claims in the Continuous-time Renewal Model Heavy Tails of Discouned Aggregae Claims in he Coninuous-ime Renewal Model Qihe Tang Deparmen of Saisics and Acuarial Science The Universiy of Iowa 24 Schae er Hall, Iowa Ciy, IA 52242, USA E-mail: qang@sa.uiowa.edu

More information

POSITIVE SOLUTIONS OF NEUTRAL DELAY DIFFERENTIAL EQUATION

POSITIVE SOLUTIONS OF NEUTRAL DELAY DIFFERENTIAL EQUATION Novi Sad J. Mah. Vol. 32, No. 2, 2002, 95-108 95 POSITIVE SOLUTIONS OF NEUTRAL DELAY DIFFERENTIAL EQUATION Hajnalka Péics 1, János Karsai 2 Absrac. We consider he scalar nonauonomous neural delay differenial

More information

A New Perturbative Approach in Nonlinear Singularity Analysis

A New Perturbative Approach in Nonlinear Singularity Analysis Journal of Mahemaics and Saisics 7 (: 49-54, ISSN 549-644 Science Publicaions A New Perurbaive Approach in Nonlinear Singulariy Analysis Ta-Leung Yee Deparmen of Mahemaics and Informaion Technology, The

More information

Heat kernel and Harnack inequality on Riemannian manifolds

Heat kernel and Harnack inequality on Riemannian manifolds Hea kernel and Harnack inequaliy on Riemannian manifolds Alexander Grigor yan UHK 11/02/2014 onens 1 Laplace operaor and hea kernel 1 2 Uniform Faber-Krahn inequaliy 3 3 Gaussian upper bounds 4 4 ean-value

More information

Solutions from Chapter 9.1 and 9.2

Solutions from Chapter 9.1 and 9.2 Soluions from Chaper 9 and 92 Secion 9 Problem # This basically boils down o an exercise in he chain rule from calculus We are looking for soluions of he form: u( x) = f( k x c) where k x R 3 and k is

More information