Markov Process Duality

Markov Process Duality

Markov Process Duality Jan M. Swart Luminy, October 23 and 24, 2013 Jan M. Swart Markov Process Duality Markov Chains S finite set. RS space of functions f : S R. ! S For probability kernel P = (P(x; y))x;y2S and f R define left and right multiplication as 2 X X Pf (x) := P(x; y)f (y) and fP(x) := f (y)P(y; x): y y (I do not distinguish row and column vectors.) Def Chain X = (Xk )k≥0 of S-valued r.v.'s is Markov chain with transition kernel P and state space S if S E f (Xk+1) (X0;:::; Xk ) = Pf (Xk ) a:s: (f R ) 2 P (X0;:::; Xk ) = (x0;:::; xk ) , = P[X0 = x0]P(x0; x1) P(xk−1; xk ): ··· µ µ µ Write P ; E for process with initial law µ = P [X0 ]. x δx x 2 · P := P with δx (y) := 1fx=yg. E similar. Jan M. Swart Markov Process Duality Markov Chains Set k µ k x µk := µP (x) = P [Xk = x] and fk := P f (x) = E [f (Xk )]: Then the forward and backward equations read µk+1 = µk P and fk+1 = Pfk : In particular π invariant law iff πP = π. Function h harmonic iff Ph = h h(Xk ) martingale. , Jan M. Swart Markov Process Duality Random mapping representation (Zk )k≥1 i.i.d. with common law ν, take values in (E; ). φ : S E S measurable E × ! P(x; y) = P[φ(x; Z1) = y]: Random mapping representation (E; ; ν; φ) always exists, highly non-unique. E X0 independent of (Zk )k≥1, then Xk := φ(Xk−1; Zk )(k 1) ≥ defines Markov chain with transition kernel P. Example if rand < 0:3 X = X + 1 else X = X 1 end − Jan M. Swart Markov Process Duality Continuous time Markov semigroup (Pt )t≥0 satisfies Ps Pt = Ps+t , limt#0 Pt = P0 = 1. Given by 1 X 1 P = e tG = tnG n; t n! n=0 where generator G satisfies G(x; y) 0 for x = y and P G(x; y) = 0. ≥ 6 y Def Process X = (Xt )t≥0 is Markov with semigroup (Pt )t≥0 and generator G if S E f (Xu) (Xs )0≤s≤t = Pu−t f (Xt ) a:s: (f R ): 2 2 P"(x; y) = 1fx=yg + "G(x; y) + O(" ) with G(x; y) jump rate. µ x µt := µPt (x) = P [Xt = x] and ft := Pt f (x) = E [f (Xt )] satisfy the forward and backward equations @ @ @t µt = µt G and @t ft = Gft : @ Also @t Pt = GPt = Pt G. Jan M. Swart Markov Process Duality Random mapping representation Write X Gf (x) = rm f (m(x)) f (x) − m2M with collection of maps m : S S and (rm)m2M nonnegative rates. M ! Let ∆ be a Poisson point subset of R with local intensity rmdt, and set M × ∆s;u := (m; t): s < t u f ≤ g =: (m1; tt );:::; (mn; tn) ; t1 < < tn: f g ··· Then Φs;u := mn m1 satisfy Φt;u Φs;t = Φs;u: ◦ · · · ◦ ◦ If X0 independent of ∆, then Xt := Φ0;t (X0)(t 0) ≥ Markov process with generator G. Jan M. Swart Markov Process Duality y I Pt = Qt , y I G = H . Duality X =(Xt )t≥0 Markov with state space S, generator G, semigroup (Pt )t≥0. Y =(Yt )t≥0 Markov with state space R, generator H, semigroup (Qt )t≥0. Def X and Y dual with duality function : S R R iff × ! x y E [ (Xt ; y)] = E [ (x; Yt )] (t 0): ≥ Implies more generally, if X and Y independent, then E[ (Xs ; Yt−s )] does not depend on s [0; t]: 2 Equivalent formulations (with Ay(x; y) := A(y; x)): X 0 0 X 0 0 I Pt (x; x ) (x ; y) = (x; y )Qt (y; y ), x0 y 0 Jan M. Swart Markov Process Duality y I G = H . Duality X =(Xt )t≥0 Markov with state space S, generator G, semigroup (Pt )t≥0. Y =(Yt )t≥0 Markov with state space R, generator H, semigroup (Qt )t≥0. Def X and Y dual with duality function : S R R iff × ! x y E [ (Xt ; y)] = E [ (x; Yt )] (t 0): ≥ Implies more generally, if X and Y independent, then E[ (Xs ; Yt−s )] does not depend on s [0; t]: 2 Equivalent formulations (with Ay(x; y) := A(y; x)): X 0 0 X 0 0 I Pt (x; x ) (x ; y) = (x; y )Qt (y; y ), x0 y 0 y I Pt = Qt , Jan M. Swart Markov Process Duality Duality X =(Xt )t≥0 Markov with state space S, generator G, semigroup (Pt )t≥0. Y =(Yt )t≥0 Markov with state space R, generator H, semigroup (Qt )t≥0. Def X and Y dual with duality function : S R R iff × ! x y E [ (Xt ; y)] = E [ (x; Yt )] (t 0): ≥ Implies more generally, if X and Y independent, then E[ (Xs ; Yt−s )] does not depend on s [0; t]: 2 Equivalent formulations (with Ay(x; y) := A(y; x)): X 0 0 X 0 0 I Pt (x; x ) (x ; y) = (x; y )Qt (y; y ), x0 y 0 y I Pt = Qt , y I G = H . Jan M. Swart Markov Process Duality Duality If the matrix is invertible, then y −1 Pt = Qt ; which relates the backward evolution of X to the forward evolution of Y . In general π invariant for Y π harmonic for X : ) y Proof Pt π = Qt π = (πQt ) = π. Similar: h harmonic for Y h invariant under right-multiplication with ) Pt (in particular, if h is a probabiity distribution, then it is an invariant law). Jan M. Swart Markov Process Duality Pathwise Duality Def Maps m : S S andm ^ : R R are dual w.r.t. if ! ! m(x); y = x; m^ (y) x; y: 8 Let X Gf (x) = rm f (m(x)) f (x) ; − m2M X Hf (y) = rm f (m ^ (y)) f (y) : − m2M Lemma For each t > 0, X and Y can be coupled such that (Xu)0≤u≤s and (Yu)0≤u≤t−s independent and Xs−; Yt−s a.s. does not depend on s [0; t]: 2 Jan M. Swart Markov Process Duality Pathwise Duality Proof Set ∆s−;u− := (m; t): s t < u f ≤ g =: (m1; tt );:::; (mn; tn) ; t1 < < tn: f g ··· Then Φ^ s−;u− :=m ^ 1 m^ n dual to Φs−;u− := mn m1: ◦ · · · ◦ ◦ · · · ◦ ^ For fixed t > 0, observe that Ys := Φ(t−s)−;t−(Y0)(s 0) Markov with generator H. Then ≥ Xs−; Yt−s = Φ0−;s−(X0); Φ^ s−;t−(Y0) = Φ0−;t−(X0); Y0 does not depend on s [0; t]. 2 Jan M. Swart Markov Process Duality A formal dual (S) := set of all subsets of S. For m : S S define mP −1 : (S) (S) by m−1(A) := x : m(!x) A inverse image. P !P f 2 g −1 Observe m dual to m w.r.t. (x; A) := 1fx2Ag: −1 (m(x); A) = 1fm(x)2Ag = 1fx2m−1(A)g = (x; m (A)): Consequence X dual to set-valued process with generator X X −1 f (A) = rm f (m (A)) f (A) : G − m2M Question Does the large ( (S) = 2jSj) space (Λ) contain any useful subspaces that are invariantjP underj the dynamicsP of ? X Jan M. Swart Markov Process Duality I x y and y x implies x = y, ≤ ≤ I x y z implies x z. ≤ ≤ ≤ I x y or y x for all x; y S; x = y. ≤ ≤ 2 6 −1 I A increasing m (A) increasing, ) −1 I A decreasing m (A) decreasing. ) Partial Order Recall that a partial order over S is a relation s.t. ≤ I x x, ≤ A partial order is a total order if Let S; S 0 be partially ordered. Then m : S S 0 is monotone if ! x y m(x) m(y): ≤ ) ≤ A set A S is increasing (decreasing) if 1A : S 0; 1 monotone (resp. ⊂ ! f g 1 1A monotone). − Observe m : S S monotone iff ! Jan M. Swart Markov Process Duality I x y z implies x z. ≤ ≤ ≤ I x y or y x for all x; y S; x = y. ≤ ≤ 2 6 −1 I A increasing m (A) increasing, ) −1 I A decreasing m (A) decreasing. ) Partial Order Recall that a partial order over S is a relation s.t. ≤ I x x, ≤ I x y and y x implies x = y, ≤ ≤ A partial order is a total order if Let S; S 0 be partially ordered. Then m : S S 0 is monotone if ! x y m(x) m(y): ≤ ) ≤ A set A S is increasing (decreasing) if 1A : S 0; 1 monotone (resp. ⊂ ! f g 1 1A monotone). − Observe m : S S monotone iff ! Jan M. Swart Markov Process Duality I x y or y x for all x; y S; x = y. ≤ ≤ 2 6 −1 I A increasing m (A) increasing, ) −1 I A decreasing m (A) decreasing. ) Partial Order Recall that a partial order over S is a relation s.t. ≤ I x x, ≤ I x y and y x implies x = y, ≤ ≤ I x y z implies x z. ≤ ≤ ≤ A partial order is a total order if Let S; S 0 be partially ordered. Then m : S S 0 is monotone if ! x y m(x) m(y): ≤ ) ≤ A set A S is increasing (decreasing) if 1A : S 0; 1 monotone (resp. ⊂ ! f g 1 1A monotone). − Observe m : S S monotone iff ! Jan M.

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    80 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us