Växjö University. Some recent simulation techniques for diffusion bridges. School of Mathematics and System Engineering

Size: px
Start display at page:

Download "Växjö University. Some recent simulation techniques for diffusion bridges. School of Mathematics and System Engineering"

Transcription

1 School of Mathematics and System Engineering Reports from MSI - Rapporter från MSI Växjö University Some recent simulation techniques for diffusion bridges Yadigar Sekerci May 2009 MSI Växjö University SE VÄXJÖ Report ISSN ISRN VXU/MSI/MA/C/ /--SE

2

3 Yadigar Sekerci Some recent simulation techniques for diffusion bridges Bachelor s thesis Mathematics 2009 Växjö University

4

5 Abstract We apply some recent numerical solutions to diffusion bridges written in Iacus (2008). One is an approximate scheme from Bladt and Sørensen (2007), another one, from Beskos et al (2006), is an algorithm which is exact: no numerical error at given grid points! Key-words: Brownian bridge, diffusion bridge, Brownian motion, stochastic differential equation, simulation, numerical methods, Euler-Maruyama s method, acceptencerejection method. Sammanfattning Vi tillämpar några nya numeriska lösningar till diffusionsbryggor beskrivna i Iacus (2008). En algoritm, från Bladt Sørensen (2007), är enkel och approximativ, den andra, från Beskos et al (2006), är exakt: inga numeriska fel vid givna gridpunkter! Nyckelord: Brownsk rörelse, diffusionsbro, Brownsk rörelse, stokastisk differentialekvation, simulering, numeriska metoder, diffusionsbro, Euler-Maruyama s method, acceptansförkastelsemetoden Acknowledgments I thank Roger Pettersson Contents 1 Introduction 1 2 Simulation of the Brownian motion 1 3 Simulation of the Brownian bridge 1 4 Simulation of stochastic differential equations 3 5 Simulation of diffusion bridges The approximate algorithm Ornstein-Uhlenbeck bridge One exact algorithm PDE-MCMC-algorithm References 11 A Appix: Definition of random variables etc 12 iii

6 B Appix: matlab codes 12 B.1 Euler-Maruyama s method for standard SDEs B.2 Brownian bridge B.3 Approximate diffusion bridge B.4 Code for simulation of Ornstein-Uhlenbeck bridge exact at grid points.. 13 iv

7 Figure 2.1: Twenty simulations of the Brownian motion where the number of gridpoints are Introduction Much of this work is inspired by nice examples of [9] where directions in the field of simulation of stochastic differential equations are presented. 2 Simulation of the Brownian motion A Brownian motion, also called Wiener process, is a continuous process {X t } t 0 starting at zero with indepent stationary increments that are normal distributed, more exactly, X t+h X t are normal distributed with zero mean and standard deviation h. One heuristic representation of the Brownian motion is that it satisfies the differential equation dx = dw, X 0 = 0 (2.1) where dw is dtn(0,1), a normal random variable with standard deviation dt! In fact the representation (2.1) is a particular case of stochastic differential equations (SDEs) often written in the form dx = µ(x)dt + σ(x)dw, X 0 = a for which there are stringent definitions in its integral form. It is simple to simulate the Brownian motion during a time interval [0,T ] at gridpoints 0 = t 0 <... < t n = T. The algorithm, with the miscellanous notation X i = X ti = X(t i ) is simply X i+1 = X i + i N(0,1), where i = ti = t i+1 t i. It means that for each time step you simulate a normal random variable with mean zero and standard deviation i. 3 Simulation of the Brownian bridge The Brownian bridge is a process {X t } that heuristically is just a Brownian motion with fixed points X 0 = a, X T = b. It satisfies the SDE dx = dw, X 0 = a, X T = b. (3.1) Note that X t is not adapted, it means X t not only deps on the information (the σ- algebra) of the Brownian motion W up till time t. X t deps on the future values. That means that even though (3.1) looks almost trivial it is actually not a standard SDE. However, it is well known that a solution to (3.1) is given by the closed formula, X t = a(1 t T ) + b t T + (W t t T W T ), 0 t T, (3.2) see any standard book, for example [2]. That makes is particularly simple to calculate the Brownain bridge: First you simulate the Brownian at grid points 0 = t 0 < t 1 <... < t n = T. Then you simulate the Brownian bridge at those grid points: X i = a(1 t i T ) + b t T + (W i t T W n), 0 t i T, (3.3) 1

8 Figure 3.1: The blue path (solid) is an exact simulation of the Brownian bridge at the grid points by the algorithm (3.3) starting and ing at zero. The green path (shadowed) is the Euler approximation (3.5) of (3.4). The number of gridpoints is The error seems to increase. For this simulation the error seems to be around the middle. When I simulated I found that the error is typically largest around the middle. It would be intersting to know a more exact description of the error. where we again used the miscellanous notation X i = X ti and W i = W ti. Now what is interesting, is that the Brownian bridge satisfies the SDE dx = b X t T t dt + dw t, 0 t T, X 0 = a. (3.4) Note that there is no -point condition X T = b. The SDE (3.4) looks like an SDE of standard type with solution X adapted to the Brownian motion, i.e. X t does not dep on future values of W. However, the drift coefficient (b X t )/(T t) is dangerously large as t approaches T. Now, an interesting question is, how bad is an Euler simulation of (3.4). In this bachelor thesis we don t reply that answer. Note that is not a trivial matter due to the non-boundedness of the drift coefficient for t near T. The Euler simulation is (again with miscellanous notation) X i+1 = X i + b X i T t i i + i N(0,1). (3.5) 4 Simulation of stochastic differential equations A generalization of the Brownian motion is a solution to the stochastic differential equation (SDE) dx t = µ(x t )dt + σ(x t )dw t, X 0 = a. (4.1) Nowadays there is a huge literature on such equations. One interpretation for tourists in this field is simply that it is the limit of the following naïve Euler-Maruyama algorithm: 2

9 X 0 = a and X i+1 = X i + µ(x i ) i + σ(x i ) i N(0,1),i = 1,2,... As before X i is X ti where t 0 = 0 < t 1 <..., i = t i+1 t i. In the next chapter on simulation of diffusion bridges, Euler-Maruyama simulation of SDEs will be used in an approximative algorithm for diffusion bridges. There are other more sophisticated methods, see e.g [9] and [5]. Throughout we assume there exists a unique solution of (4.1). Classical assumptions that give existence and uniqueness of a solution to (4.1) is linear growth and a Lipschitz continuity of b and σ 1. 5 Simulation of diffusion bridges A generalization of (2.1) is dx t = µ(x t )dt + σ(x t )dw t, 0 < t < T, X(0) = a, X(T ) = b (5.1) Given µ and σ it is often convenient to call a solution to (5.1) a (0,a,T,b)-bridge. It is not a trivial matter how to simulate (5.1). In the literature it seems that one important reason why diffusion bridge simulation is needed is that it is used for likelihood inference of discretely observed diffusion processes. That means that simulation of diffusion bridges is used for estimating parameters of SDEs observed at some grid points. In [3] it is suggested a particular technique based on Markov Chain Monte Carlo (MCMC) methods. In [7] there is a technique emaneting from the observation that limit solutions of certain stochastic partial differential equations are solutions to diffusion bridges! In [4] it is suggested a technique which is exact at any given time points the time interval [0,T ]! In this thesis we apply the approximate technique from [8] and the technique from [4]. 5.1 The approximate algorithm In [8] there is a simple algorithm to a simulate an approximate solution to (5.1), it means an approximate diffusion bridge. For the algorithm to work, time-reversibility of the diffusion process {X t [0,T ] } is assumed, that means {X t X 0 : t [0,T ]} and the process {X T t X T : t [0,T ]}, have the same statistical properties: the process behaves similarly if the time is running backwards. We impose the following condition, which ensures that the diffusion X is time-reversible, see [8], [5]. For that condition we let (l,r), l < r be the state space of {X t [0,T ] }, that means l is the smallest possible value for {X t [0,T ] } which may be and r is the largest possible value for {X t [0,T ] } which may be. Condition 5.1. Let {X t [0,T ] } be the solution to the SDE (4.1). Let x µ(y) s(x) = exp( 2 x# σ 2 (y) dy) where x# denotes an arbitrary point in the state space (l,r). Let m(x) = 1 σ 2 (x)s(x). 1 f Lipschitz continuous means f (x) f (y) L x y for all x,y. 3

10 Figure 5.1: The figures describes the approximate algorithm for the diffusion bridge dx = 0.5Xdt + dw, X 0 = 0, X T = 1. In the upper figure the blue one (solid line) starts at 0, the green one (dotted) starts at 1 (!) and the red one (dashed) is the flipped version of the green one, it s at 1. The bridge starts with the blue (solid) one and if it crosses the red (dashed) one the bridge continues with the red one which finishes at 1. Assume and x x# r l s(x)dx = m(x)dx < x# l s(x)dx =. (5.2) The approximative algorithm from [8] for simulation of diffusion bridges is as follows. Let W 1 and W 2 be two indepent Wiener processes, and let X 1 and X 2 as the solutions to dxt 1 = µ(xt 1 )dt + σ(xt 1 )dwt 1, X0 1 = a and dxt 2 = µ(xt 2 )dt + σ(xt 2 )dwt 2, X0 2 = b. The idea is to use the time-reversibility of X 2 and realize a (0,a,T,b) -bridge by simulating the process X 1 from a forward in time and X 2 from b backward in time! If the samples paths of the two process intersect, they can be combined into a realization of the bridge. Fact 5.1 ([8]). Let τ = inf(0 t 1Xt 1 = X1 t 2 ), where the infimum over empty set is taken to be infinite. Define { X 1 Z t = t, if 0 t τ X1 t 2 if τ < t 1 Assume Condition 5.1. Then, the distribution of {Z t : 0 t T } conditional on the event {τ T } is equal to the conditional distribution of {X t : 0 t T } given X 0 = a and X T = b; i.e., Z is a (0,a,T,b)-diffusion bridge. The fact 5.1 points towards a natural algorithm for simulation of a diffusion bridge.the algorithm consists in simulating two indepent diffusion process X 1 and X 2 using one numerical method, for example the Euler-Maruyama on the time interval [0, T ] with discretization step = T /n applying a rejection sampling procedure. Let Yi 1 and Y i 2, i = 0,1,2,...,n be indepent simulations of X 1 and X 2. If either Y 1 i Y 2 Y 1 (i+1) Y 2 (n (i+1)) or Y 1 i Y 2 (n i) and Y 1 (i+1) Y 2 (n (i+1)), a crossing has been realized. Hence, let ν = min{i (1,2,...,n) Yi 1 Y (n i) 2 } if Y 0 1 Y T 2 (1,2,...,n) Yi 1 Y (n i) 2 } if Y 0 1 Y T 2, and define { Y 1 B i = i, if i = 0,1,...,ν 1 if i = ν,...,n Y 2 (n i) (n i) and and v = min{i Then B is simulation of a (0,a,T,b)-diffusion bridge. If no crossing happened,start again by simulating Yi 1 and Y i 2 and iterate until a crossing of the two trajectories is realized. 4

11 Figure 5.2: 20 simulated OU-bridges with the algorithm exact at the grid points. Figure 5.3: 20 simulated OU-bridges with the approximate algorithm 5.2 Ornstein-Uhlenbeck bridge For the Ornstein-Uhlenbeck bridge (OU-bridge) dx t = θx t dt + σdw t, X 0 = a, X T = b (5.3) there is a simple algorithm that is exact at the grid points, [8]. According to [8], from the transition densities of the Ornstein-Uhlenbeck process we can calculate the transition densities of the OU-bridge. Thus we could in principle simulate the OU-bridge by sampling transitions from these densities. The following alternative method is, however, numerically more stable. For this particular equation there exists an algorithm which at the grid points t 1,t 2 is exactly an OU-bridge. That means that at the grid points the simulated process has the same distribution as the OU-bridge. The algorithm from the lemma below is from [8]: Lemma 5.1. For a partition 0 = t 0 < t 1 <... < t n = T, define X i = X ti and Y i = Y ti by X 0 = a, X i+1 = e θ(t i+1 t i ) X ti +Y i, i = 1,...n where Y i+1 = ( σ 2 2θ (1 e 2θ(t i+1 t i ))) 1/2 N(0,1), i = 1,...n where N(0,1) represents indep N(0,1) random variables. Let Z i be defined by Z i = X i + (b X n ) eθt i e θt i e θt n e θt n,, i = 1,...n. Then Z ti = Z i has the same distribution as a Ornstein-Uhlenbeck bridge at the grid points 0 = t 0 < t 1 <... < t n = T, with Z 0 = a, Z T = b. Lemma 5.1 makes it possible to compare the approximate algorithm with the OUalgorithm exact at the grid points. See figures Figure 5.4: Kernel density estimate of the Ornstein-Uhlenbeck bridge at midpoint time t = 1/2 with the algorithm exact at the grid points and the kernel density estimate of the OU-bridge with the approximate algorithm. For the approximate algorithm the number of time steps is 500 and the number of simulations is = For the exact algorithm only the mid-point time is simulated (because the random draw is exact) and the number of simulations is also , due to slowness of the exact algorithm. 5

12 5.3 One exact algorithm Let us first describe an algorithm for SDEs (4.1). We later describe a modified algorithm for diffusion bridges (5.1). In [4] is suggested a general technique for simulation of SDE (4.1) for which the simulated process has the same distribution as the process given by (4.1). It is not, as for example, the Euler-Maruyama method only an approximative algorithm. The conditions we apply here is the following condition. Condition 5.2. (i) The derivative µ x of µ exists (ii) There exists k 1 and k 2 such that k µ2 (x) µ x(x) k 2 for any x R Condition 5.2 can be relaxed but then the algorithm is more technical. In the algorithm below we use the function φ(x) = 2 1µ2 (x)+ 2 1 µ x(x) k 1. Note that 0 φ(x) M = k 2 k 1 The exact algorithm for simulating (4.1) is briefly as follows. For a partition {0 = t 0 < t 1 < < t n = T }, given that X ti is already simulated, X ti+1 is first simulated in a certain way, typically by a acceptance-rejection rule (not a variant of the Euler-Maruyama method). To simulate the process in [t i,t i+1 ], a Poisson random number (an integer) is simulated. If that Poisson random integer, N say, is positive, N couples of uniformly random variables {(τ 1,v 1 ),...,(τ N,v N )}, where τ 1,...,τ N are uniform random numbers in [t i,t i+1 ] and (v 1,...,v N ) are uniform random in the set [0,M]. Then simulate a standard Brownian bridge in [t i,t i+1 ] beginning at X ti and ing at X ti+1. Consider the values y 1,...y N of the Brownian bridge at the time points τ 1,...,τ N. The Brownian bridge is accepted as a simulation of {X t : t [t i,t i+1 ]} if φ(y 1 ) v 1,...,φ(y N ) v N. If we only want values of {X t } at the grid points 0 = t 1 < t 2 <... < t n = T, the connecting Brownian bridges are not needed. An alternative would instead be to only simulate the point X T exactly by an acceptance-rejection rule, and then simulate the Brownian bridge in the whole interval [0,T ] and, if the Brownian bridge is accepted, the values of the Brownian bridge at the grid points is obtained as simulated values of the process {X t : t [0,T ]}. However, the rejection probability of the X T -value and the Brownian bridge is in that situation probably high which should imply an inefficient algorithm. For diffusion bridges of type (5.1) the exact algorithm is simpler. Given X 0 = a, X T does not need to be simulated by a time consuming acceptance-rejection rule: X T is just given the value b. Instead the skeleton time points τ 1,...,τ k can be uniformly picked in the whole time interval [0,T ]. It is applied to diffusion bridges of the type dx t = µ(x t )dt + dw t, X 0 = a, X T = b, (5.4) i.e. σ 1. Note that (5.4) is more general than one may think at the first sight. The equation (5.1) with possibly non-constant σ can, under non-degeneracity conditions of σ (for example σ(x) > ε > 0 for any x) be transformed into one with diffusion coefficient one by applying the Lamperti transform, Xt 1 Y t = F(X t ) = z σ(u) du. Here z is any arbitrary value in the state space of {X t }. Indeed, the process {Y t } solves the SDE dy t = µ Y (Y t )dt + dw t, where µ Y (Y t ) = µ(f 1 (y)) σ(f 1 (y) 1 2 σ x(f 1 (y)) 6

13 Figure 5.5: 20 simulated bridges of dx = sin(x)dt +dw, X 0 = X 1 = 0 with the algorithm exact at the grid points. Figure 5.6: 20 simulated bridges of dx = sin(x)dt + dw, X 0 = X 1 = 0 with the approximate algorithm which can also be written as dy t = ( µ(x t) σ(x t ) 1 2 σ x(x t ))dt + dw t. The exact algorithm to simulate the diffusion bridge (5.4), from [4], see also [9]: 1. Let X T = b. 2. Simulate a non-negative integer τ = N from the Poisson distribution with intensity λ = T M. 3. Draw (if N 1) k pairs of uniformly distributed random numbers {(τ i,v i ) : i = 1,...,N} on [0,T ] [0,M],i=1,2,...,N. 4. Generate a Brownian bridge X starting at x at time 0 and ing in b at time T at time instants t i ; i.e., generate X τi = y i, i = 1,...,N. 5. Compute the indicator function I = N i=1 1 {φ(yi ) ν i }. 6. If I = 1, the trajectory {Y 0 = a,y t1 = y 1,...,Y tk = y k,y T = b} is accepted as an exact draw of a solution X to (5.4) at the points 0 < t 1 <... < t k < T. Otherwise, restart from step 1. Exact draws from X given by (5.4) at pre-determined grid-points 0 = t 1 < t 2 <... < t n = T is obtained by noting the values at these points of the connecting Brownian bridges between the points τ 1 <... < τ k. Figure 5.7: Kernel density estimate of the bridge dx = sin(x)dt + dw, X 0 = X 1 = 1 at midpoint time t = 1/2 with the algorithm exact at the grid points and the kernel density estimate of the bridge with the approximate algorithm. For the exact algorithm the number of time steps is only 3 (we only need the midpoint!) and the number of simulations is 500. For the approximate algorithm the number of time steps is 501 and the number of simulations only 500. In this situation I don t know a closed formula for the density of X 1/2, the midpoint value. 7

14 Figure 5.8: Simulation of a numerical solution to (5.6),[6]. The solution to (5.6) converges to a diffusion bridge. 5.4 PDE-MCMC-algorithm From [7], for X given by the diffusion bridge SDE dx(x) = µ(x(x))dx + σdw(x), X(0) = X(1) = 0 (5.5) (here the time variable is x; the reason will be clear from the arguments below) can be sampled by MCMC-sampling of u given by stochastic partial differential equation u t = 1 ( 2 u µ(u)µ (u) 1 σ 2 x 2 2 µ (u)) W t x u(0,t) = u(1,t) = 0, t [0,T ] u(x,0) = u 0 (x), x [0,1]. (5.6) where 2 W t x is a time-space Gaussian white noise. MCMC (Markov Chain Monte Carlo) is a particular class of algorithms for sampling from probability distributions based on constructing a Markov chain that has the desired distribution as its equilibrium distribution. The state of the chain after a large number of steps is then used as a sample from the desired distribution. The quality of the sample improves as a function of the number of steps. What is used here is that (for a fixed x [0,1], as t, u(x,t) converges to a process X(x) which solves (5.5). What is needed is to simulate u(t,x) for large t in a clever way. Normally when simulations are needed the time interval is bounded, but here the focus is on for large times. For that the, for me complicated MCMC-algorithm, is needed. References [1] S: Iacus, Simulation and inference for stochastic differential equations, with R examples, Springer, (2008). [2], I Karatzas, I Shreve, Brownian motion and stochastic calculus, Springer (1991). [3] G. Roberts, O. Stramer, On inference for partial observations from partial observed nonlinear diffusion model using Metropolis-Hastings algorithms, Biometrika 88:3 (2001), [4] A.Beskos, O.Papaspiliopoulos and G.O. Roberts, Retrospective exact simulation of diffusion sample paths with applications, Bernoulli 12:6 (2006), [5] J. Kent, Time-reversible diffusions, Adv Appl Probab 10 (1978), [6] R. Pettersson, Projection scheme for a reflected stochastic heat equation with additove noise, Foundations of Probability and Physics-3, AIP Conference Proceedings, ed: A. Yu Khrennikov, (2005), [7] A.M. Stuart, J. Voss, and P. Wiberg Conditional path sampling of SDEs and the Langevin MCMC method. Comm. Math. Sci. 2 (2004),

15 [8] M.Bladt and M.Sorensen, Simple simulation of diffusion bridges with application to likelihood inference for diffusions,centre for Analytical Finance, Working Paper, University of Copenhagen. February (2007) [9] S. M. Iacus, Simulation and interference for Stochastic Differential Equations, with R Examples, Springer Series in Statistics, (2008) A Appix: Definition of random variables etc A random variable X is a function Ω ω X(ω) measurable in a certain sense so that probabilities can be defined for sets of the type {X(ω) x}. The ω represents a random choice. A real-valued random variable simply takes its values in R. A continuous stochastic proces X = {X t : t 0} is a random variable with values in the set of continuous functions C(R). The map then is Ω = C(R) X t (ω) = ω(t), i.e. the random selection is then a continuous function. A Brownian motion is a continuous stochastic process {W t : t 0}, W(0) = 0, with indepent increments W t+h W t that are hn(0,1) for h > 0. An Ornstein-Uhlenbeck process is a process {X t 0 } satisfying dx t = θx t dt + σdw t, X 0 = a i.e. as the OU-bridge (5.3) but without restrictions on X T. B Appix: matlab codes B.1 Euler-Maruyama s method for standard SDEs Matlab code for Euler simulation of SDE of the type dx = µ(x)dt + σ(x)dw, X 0 = x. In the code below mu(x) = sin(x), σ(x) = 1. x0=1; % initial value n=1000; % number of partition points t=linspace (0,1,n); %gridpoints dt=1/(n-1); %partition step N=500; % N number of simulations x=x0*ones (1,N); mu=@(x) sin(x); sigma=@(x)1; dw=sqrt (dt)*randn (n,n); for i=1:n-1; x (i+1,:)=x (i,:)+mu(x (i,:))*dt+sigma (x (i,:)).*dw (i,:); B.2 Brownian bridge Matlab code for Brownian bridge X t = a(1 t T ) + b t T + (W t t T W T ), 0 t T, 9

16 n=10000; % number of partition points T=1; t=linspace(0,t,n) ; dt=1/(n-1); N=5; dw=sqrt (dt)*randn (n,n); W=cumsum(dW); a=0; b=0; X=a*(1-t/T)+b*t/T+(W-t/T.*W(,:) ); B.3 Approximate diffusion bridge The matlab code for the approximate diffusion bridge that appeared in [8]. n=501; %number of partition points N=500; %number of simulations t=linspace(0,t,n) ; ok=0; dt=t(2)-t(1); bridges=zeros(n,n); for j=1:n ok=0; while ok==0; x=a;y=b; z=a; dwx=sqrt (dt)*randn (n,1); dwy=sqrt (dt)*randn (n,1); for i=1:n-1; x (i+1)=x (i)+mu(x (i))*dt+sigma (x (i)).*dwx (i); y (i+1)=y (i)+mu(y (i))*dt+sigma (y (i)).*dwy (i); z=fliplr(y); dummy=(x(1:-1)>=z(1:-1)).*(x(2:)<=z(2:))... (x(1:-1)<=z(1:-1)).*(x(2:)>=z(2:)); if max(dummy)==1 ok=1; %they meet index=min(find(dummy==1)); bridges(:,j)=[x(1:max(index-1,1)) z(max(index,2):)]; B.4 Code for simulation of Ornstein-Uhlenbeck bridge exact at grid points Matlab code for simulation of exact at the grid points, [8]. dx = θ(x t )dt + σ(dw t ), X 0 = a, X T = b n=100; % number of partition points N=n^2; %number of simulations 10

17 T=1; t=linspace(0,t,n) ; %grid points dt=1/(n-1); a=0; b=1; theta=.5; sig=1; Y=(sig/sqrt(2*theta)*sqrt(1-exp(-2*theta*diff(t)))*ones(1,N)).*... randn(n-1,n); X=a*ones(1,N); for i=1:n-1 X(i+1,:)=exp(-theta*dt)*X(i,:)+Y(i,:); Z=X+(exp(theta*t)-exp(-theta*t))/(exp(theta*t(n))-... exp(-theta*t(n))*(b-x(n,:)); Algorithm for more general diffusion bridges exact at grid points Matlab code for dx t = µ(x t )dt + dw t, X 0 = a,x T = b, i.e. σ 1, where k µ2 (x) µ x(x) k 2 N=500;%number of simulations n=3;%number of partition points mu=@(x)sin(x); mup=@(x)cos(x); sigma=@(x)1; a=0; b=0; T=1;%final time prephi=@(x)1/2*mu(x).^2+1/2*mup(x); mprephi=@(x)-prephi(x); k1=prephi(fminsearch(prephi,-pi));%-1/2; k2=prephi(fminsearch(mprephi,pi/2));%5/8; phi=@(x)prephi(x)-k1; M=k2-k1; lambda=t*m; Ndifb=zeros(n,N); t=linspace(0,t,n) ; for j=1:n accept=0; while accept==0 k=poissrnd(lambda); if k>0 pretskel=sort(t*rand(k,1)); v=m*rand(k,1); tskel=[0; pretskel; T]; W=[0; cumsum(sqrt(diff(tskel)).*randn(length(tskel)-1,1))]; skeleton=a+w-tskel/t*(w()-b+a); I=sum((phi(skeleton(2:-1))<=v));%prod 11

18 if I==k accept=1; % [tskel skeleton] difb=a; for i=2:length(skeleton) if max((t>tskel(i-1)&t<tskel(i)))>0 s=t(t>tskel(i-1)&t<tskel(i)); difb=[difb; Bbfn(tskel(i-1),skeleton(i-1),tskel(i),... difb=[difb;b]; Ndifb(:,j)=difb; %plot(t,difb) %plot(ti,x); 12

19

20 Växjö universitet Matematiska och systemtekniska institutionen SE Växjö Tel. +46 (0) , fax +46 (0)

Exact Simulation of Diffusions and Jump Diffusions

Exact Simulation of Diffusions and Jump Diffusions Exact Simulation of Diffusions and Jump Diffusions A work by: Prof. Gareth O. Roberts Dr. Alexandros Beskos Dr. Omiros Papaspiliopoulos Dr. Bruno Casella 28 th May, 2008 Content 1 Exact Algorithm Construction

More information

Lecture 1: Pragmatic Introduction to Stochastic Differential Equations

Lecture 1: Pragmatic Introduction to Stochastic Differential Equations Lecture 1: Pragmatic Introduction to Stochastic Differential Equations Simo Särkkä Aalto University, Finland (visiting at Oxford University, UK) November 13, 2013 Simo Särkkä (Aalto) Lecture 1: Pragmatic

More information

arxiv: v1 [math.pr] 1 Jul 2013

arxiv: v1 [math.pr] 1 Jul 2013 ESTIMATION OF FIRST PASSAGE TIME DENSITIES OF DIFFUSIONS PROCESSES THROUGH TIME-VARYING BOUNDARIES arxiv:307.0336v [math.pr] Jul 03 Imene Allab and Francois Watier Department of Mathematics, Université

More information

Gaussian processes for inference in stochastic differential equations

Gaussian processes for inference in stochastic differential equations Gaussian processes for inference in stochastic differential equations Manfred Opper, AI group, TU Berlin November 6, 2017 Manfred Opper, AI group, TU Berlin (TU Berlin) inference in SDE November 6, 2017

More information

Lecture 4: Introduction to stochastic processes and stochastic calculus

Lecture 4: Introduction to stochastic processes and stochastic calculus Lecture 4: Introduction to stochastic processes and stochastic calculus Cédric Archambeau Centre for Computational Statistics and Machine Learning Department of Computer Science University College London

More information

Stochastic differential equation models in biology Susanne Ditlevsen

Stochastic differential equation models in biology Susanne Ditlevsen Stochastic differential equation models in biology Susanne Ditlevsen Introduction This chapter is concerned with continuous time processes, which are often modeled as a system of ordinary differential

More information

Introduction to multiscale modeling and simulation. Explicit methods for ODEs : forward Euler. y n+1 = y n + tf(y n ) dy dt = f(y), y(0) = y 0

Introduction to multiscale modeling and simulation. Explicit methods for ODEs : forward Euler. y n+1 = y n + tf(y n ) dy dt = f(y), y(0) = y 0 Introduction to multiscale modeling and simulation Lecture 5 Numerical methods for ODEs, SDEs and PDEs The need for multiscale methods Two generic frameworks for multiscale computation Explicit methods

More information

TMS165/MSA350 Stochastic Calculus, Lecture on Applications

TMS165/MSA350 Stochastic Calculus, Lecture on Applications TMS165/MSA35 Stochastic Calculus, Lecture on Applications In this lecture we demonstrate how statistical methods such as the maximum likelihood method likelihood ratio estimation can be applied to the

More information

Homogenization with stochastic differential equations

Homogenization with stochastic differential equations Homogenization with stochastic differential equations Scott Hottovy shottovy@math.arizona.edu University of Arizona Program in Applied Mathematics October 12, 2011 Modeling with SDE Use SDE to model system

More information

Exact Simulation of Multivariate Itô Diffusions

Exact Simulation of Multivariate Itô Diffusions Exact Simulation of Multivariate Itô Diffusions Jose Blanchet Joint work with Fan Zhang Columbia and Stanford July 7, 2017 Jose Blanchet (Columbia/Stanford) Exact Simulation of Diffusions July 7, 2017

More information

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS PROBABILITY: LIMIT THEOREMS II, SPRING 218. HOMEWORK PROBLEMS PROF. YURI BAKHTIN Instructions. You are allowed to work on solutions in groups, but you are required to write up solutions on your own. Please

More information

Simulation Method for Solving Stochastic Differential Equations with Constant Diffusion Coefficients

Simulation Method for Solving Stochastic Differential Equations with Constant Diffusion Coefficients Journal of mathematics and computer Science 8 (2014) 28-32 Simulation Method for Solving Stochastic Differential Equations with Constant Diffusion Coefficients Behrouz Fathi Vajargah Department of statistics,

More information

Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio ( )

Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio ( ) Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio (2014-2015) Etienne Tanré - Olivier Faugeras INRIA - Team Tosca November 26th, 2014 E. Tanré (INRIA - Team Tosca) Mathematical

More information

Brownian motion. Samy Tindel. Purdue University. Probability Theory 2 - MA 539

Brownian motion. Samy Tindel. Purdue University. Probability Theory 2 - MA 539 Brownian motion Samy Tindel Purdue University Probability Theory 2 - MA 539 Mostly taken from Brownian Motion and Stochastic Calculus by I. Karatzas and S. Shreve Samy T. Brownian motion Probability Theory

More information

Brownian Motion. 1 Definition Brownian Motion Wiener measure... 3

Brownian Motion. 1 Definition Brownian Motion Wiener measure... 3 Brownian Motion Contents 1 Definition 2 1.1 Brownian Motion................................. 2 1.2 Wiener measure.................................. 3 2 Construction 4 2.1 Gaussian process.................................

More information

Nonparametric Drift Estimation for Stochastic Differential Equations

Nonparametric Drift Estimation for Stochastic Differential Equations Nonparametric Drift Estimation for Stochastic Differential Equations Gareth Roberts 1 Department of Statistics University of Warwick Brazilian Bayesian meeting, March 2010 Joint work with O. Papaspiliopoulos,

More information

Least Squares Estimators for Stochastic Differential Equations Driven by Small Lévy Noises

Least Squares Estimators for Stochastic Differential Equations Driven by Small Lévy Noises Least Squares Estimators for Stochastic Differential Equations Driven by Small Lévy Noises Hongwei Long* Department of Mathematical Sciences, Florida Atlantic University, Boca Raton Florida 33431-991,

More information

Introduction to numerical simulations for Stochastic ODEs

Introduction to numerical simulations for Stochastic ODEs Introduction to numerical simulations for Stochastic ODEs Xingye Kan Illinois Institute of Technology Department of Applied Mathematics Chicago, IL 60616 August 9, 2010 Outline 1 Preliminaries 2 Numerical

More information

Stochastic Volatility and Correction to the Heat Equation

Stochastic Volatility and Correction to the Heat Equation Stochastic Volatility and Correction to the Heat Equation Jean-Pierre Fouque, George Papanicolaou and Ronnie Sircar Abstract. From a probabilist s point of view the Twentieth Century has been a century

More information

I forgot to mention last time: in the Ito formula for two standard processes, putting

I forgot to mention last time: in the Ito formula for two standard processes, putting I forgot to mention last time: in the Ito formula for two standard processes, putting dx t = a t dt + b t db t dy t = α t dt + β t db t, and taking f(x, y = xy, one has f x = y, f y = x, and f xx = f yy

More information

Simulation of conditional diffusions via forward-reverse stochastic representations

Simulation of conditional diffusions via forward-reverse stochastic representations Weierstrass Institute for Applied Analysis and Stochastics Simulation of conditional diffusions via forward-reverse stochastic representations Christian Bayer and John Schoenmakers Numerical methods for

More information

Introduction to Diffusion Processes.

Introduction to Diffusion Processes. Introduction to Diffusion Processes. Arka P. Ghosh Department of Statistics Iowa State University Ames, IA 511-121 apghosh@iastate.edu (515) 294-7851. February 1, 21 Abstract In this section we describe

More information

Markov Chain Monte Carlo (MCMC)

Markov Chain Monte Carlo (MCMC) Markov Chain Monte Carlo (MCMC Dependent Sampling Suppose we wish to sample from a density π, and we can evaluate π as a function but have no means to directly generate a sample. Rejection sampling can

More information

LAN property for ergodic jump-diffusion processes with discrete observations

LAN property for ergodic jump-diffusion processes with discrete observations LAN property for ergodic jump-diffusion processes with discrete observations Eulalia Nualart (Universitat Pompeu Fabra, Barcelona) joint work with Arturo Kohatsu-Higa (Ritsumeikan University, Japan) &

More information

Discretization of SDEs: Euler Methods and Beyond

Discretization of SDEs: Euler Methods and Beyond Discretization of SDEs: Euler Methods and Beyond 09-26-2006 / PRisMa 2006 Workshop Outline Introduction 1 Introduction Motivation Stochastic Differential Equations 2 The Time Discretization of SDEs Monte-Carlo

More information

Harnack Inequalities and Applications for Stochastic Equations

Harnack Inequalities and Applications for Stochastic Equations p. 1/32 Harnack Inequalities and Applications for Stochastic Equations PhD Thesis Defense Shun-Xiang Ouyang Under the Supervision of Prof. Michael Röckner & Prof. Feng-Yu Wang March 6, 29 p. 2/32 Outline

More information

Interest Rate Models:

Interest Rate Models: 1/17 Interest Rate Models: from Parametric Statistics to Infinite Dimensional Stochastic Analysis René Carmona Bendheim Center for Finance ORFE & PACM, Princeton University email: rcarmna@princeton.edu

More information

Bayesian inference for stochastic differential mixed effects models - initial steps

Bayesian inference for stochastic differential mixed effects models - initial steps Bayesian inference for stochastic differential ixed effects odels - initial steps Gavin Whitaker 2nd May 2012 Supervisors: RJB and AG Outline Mixed Effects Stochastic Differential Equations (SDEs) Bayesian

More information

Asymptotical distribution free test for parameter change in a diffusion model (joint work with Y. Nishiyama) Ilia Negri

Asymptotical distribution free test for parameter change in a diffusion model (joint work with Y. Nishiyama) Ilia Negri Asymptotical distribution free test for parameter change in a diffusion model (joint work with Y. Nishiyama) Ilia Negri University of Bergamo (Italy) ilia.negri@unibg.it SAPS VIII, Le Mans 21-24 March,

More information

p 1 ( Y p dp) 1/p ( X p dp) 1 1 p

p 1 ( Y p dp) 1/p ( X p dp) 1 1 p Doob s inequality Let X(t) be a right continuous submartingale with respect to F(t), t 1 P(sup s t X(s) λ) 1 λ {sup s t X(s) λ} X + (t)dp 2 For 1 < p

More information

Malliavin Calculus in Finance

Malliavin Calculus in Finance Malliavin Calculus in Finance Peter K. Friz 1 Greeks and the logarithmic derivative trick Model an underlying assent by a Markov process with values in R m with dynamics described by the SDE dx t = b(x

More information

Kolmogorov Equations and Markov Processes

Kolmogorov Equations and Markov Processes Kolmogorov Equations and Markov Processes May 3, 013 1 Transition measures and functions Consider a stochastic process {X(t)} t 0 whose state space is a product of intervals contained in R n. We define

More information

Introduction to Machine Learning CMU-10701

Introduction to Machine Learning CMU-10701 Introduction to Machine Learning CMU-10701 Markov Chain Monte Carlo Methods Barnabás Póczos & Aarti Singh Contents Markov Chain Monte Carlo Methods Goal & Motivation Sampling Rejection Importance Markov

More information

Stochastic Gradient Descent in Continuous Time

Stochastic Gradient Descent in Continuous Time Stochastic Gradient Descent in Continuous Time Justin Sirignano University of Illinois at Urbana Champaign with Konstantinos Spiliopoulos (Boston University) 1 / 27 We consider a diffusion X t X = R m

More information

Lecture 21: Stochastic Differential Equations

Lecture 21: Stochastic Differential Equations : Stochastic Differential Equations In this lecture, we study stochastic differential equations. See Chapter 9 of [3] for a thorough treatment of the materials in this section. 1. Stochastic differential

More information

Week 9 Generators, duality, change of measure

Week 9 Generators, duality, change of measure Week 9 Generators, duality, change of measure Jonathan Goodman November 18, 013 1 Generators This section describes a common abstract way to describe many of the differential equations related to Markov

More information

Stochastic optimal control with rough paths

Stochastic optimal control with rough paths Stochastic optimal control with rough paths Paul Gassiat TU Berlin Stochastic processes and their statistics in Finance, Okinawa, October 28, 2013 Joint work with Joscha Diehl and Peter Friz Introduction

More information

Controlled Diffusions and Hamilton-Jacobi Bellman Equations

Controlled Diffusions and Hamilton-Jacobi Bellman Equations Controlled Diffusions and Hamilton-Jacobi Bellman Equations Emo Todorov Applied Mathematics and Computer Science & Engineering University of Washington Winter 2014 Emo Todorov (UW) AMATH/CSE 579, Winter

More information

LAN property for sde s with additive fractional noise and continuous time observation

LAN property for sde s with additive fractional noise and continuous time observation LAN property for sde s with additive fractional noise and continuous time observation Eulalia Nualart (Universitat Pompeu Fabra, Barcelona) joint work with Samy Tindel (Purdue University) Vlad s 6th birthday,

More information

Monte Carlo methods for sampling-based Stochastic Optimization

Monte Carlo methods for sampling-based Stochastic Optimization Monte Carlo methods for sampling-based Stochastic Optimization Gersende FORT LTCI CNRS & Telecom ParisTech Paris, France Joint works with B. Jourdain, T. Lelièvre, G. Stoltz from ENPC and E. Kuhn from

More information

Sequential maximum likelihood estimation for reflected Ornstein-Uhlenbeck processes

Sequential maximum likelihood estimation for reflected Ornstein-Uhlenbeck processes Sequential maximum likelihood estimation for reflected Ornstein-Uhlenbeck processes Chihoon Lee Department of Statistics Colorado State University Fort Collins, CO 8523 Jaya P. N. Bishwal Department of

More information

A new approach for investment performance measurement. 3rd WCMF, Santa Barbara November 2009

A new approach for investment performance measurement. 3rd WCMF, Santa Barbara November 2009 A new approach for investment performance measurement 3rd WCMF, Santa Barbara November 2009 Thaleia Zariphopoulou University of Oxford, Oxford-Man Institute and The University of Texas at Austin 1 Performance

More information

Basic Definitions: Indexed Collections and Random Functions

Basic Definitions: Indexed Collections and Random Functions Chapter 1 Basic Definitions: Indexed Collections and Random Functions Section 1.1 introduces stochastic processes as indexed collections of random variables. Section 1.2 builds the necessary machinery

More information

Rough Burgers-like equations with multiplicative noise

Rough Burgers-like equations with multiplicative noise Rough Burgers-like equations with multiplicative noise Martin Hairer Hendrik Weber Mathematics Institute University of Warwick Bielefeld, 3.11.21 Burgers-like equation Aim: Existence/Uniqueness for du

More information

Stochastic Modelling in Climate Science

Stochastic Modelling in Climate Science Stochastic Modelling in Climate Science David Kelly Mathematics Department UNC Chapel Hill dtbkelly@gmail.com November 16, 2013 David Kelly (UNC) Stochastic Climate November 16, 2013 1 / 36 Why use stochastic

More information

A Class of Fractional Stochastic Differential Equations

A Class of Fractional Stochastic Differential Equations Vietnam Journal of Mathematics 36:38) 71 79 Vietnam Journal of MATHEMATICS VAST 8 A Class of Fractional Stochastic Differential Equations Nguyen Tien Dung Department of Mathematics, Vietnam National University,

More information

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS PROBABILITY: LIMIT THEOREMS II, SPRING 15. HOMEWORK PROBLEMS PROF. YURI BAKHTIN Instructions. You are allowed to work on solutions in groups, but you are required to write up solutions on your own. Please

More information

Lecture 12: Detailed balance and Eigenfunction methods

Lecture 12: Detailed balance and Eigenfunction methods Miranda Holmes-Cerfon Applied Stochastic Analysis, Spring 2015 Lecture 12: Detailed balance and Eigenfunction methods Readings Recommended: Pavliotis [2014] 4.5-4.7 (eigenfunction methods and reversibility),

More information

Simulation methods for stochastic models in chemistry

Simulation methods for stochastic models in chemistry Simulation methods for stochastic models in chemistry David F. Anderson anderson@math.wisc.edu Department of Mathematics University of Wisconsin - Madison SIAM: Barcelona June 4th, 21 Overview 1. Notation

More information

A Short Introduction to Diffusion Processes and Ito Calculus

A Short Introduction to Diffusion Processes and Ito Calculus A Short Introduction to Diffusion Processes and Ito Calculus Cédric Archambeau University College, London Center for Computational Statistics and Machine Learning c.archambeau@cs.ucl.ac.uk January 24,

More information

Partial Differential Equations with Applications to Finance Seminar 1: Proving and applying Dynkin s formula

Partial Differential Equations with Applications to Finance Seminar 1: Proving and applying Dynkin s formula Partial Differential Equations with Applications to Finance Seminar 1: Proving and applying Dynkin s formula Group 4: Bertan Yilmaz, Richard Oti-Aboagye and Di Liu May, 15 Chapter 1 Proving Dynkin s formula

More information

A stochastic particle system for the Burgers equation.

A stochastic particle system for the Burgers equation. A stochastic particle system for the Burgers equation. Alexei Novikov Department of Mathematics Penn State University with Gautam Iyer (Carnegie Mellon) supported by NSF Burgers equation t u t + u x u

More information

Quantization of stochastic processes with applications on Euler-Maruyama schemes

Quantization of stochastic processes with applications on Euler-Maruyama schemes UPTEC F15 063 Examensarbete 30 hp Oktober 2015 Quantization of stochastic processes with applications on Euler-Maruyama schemes Viktor Edward Abstract Quantization of stochastic processes with applications

More information

Universal examples. Chapter The Bernoulli process

Universal examples. Chapter The Bernoulli process Chapter 1 Universal examples 1.1 The Bernoulli process First description: Bernoulli random variables Y i for i = 1, 2, 3,... independent with P [Y i = 1] = p and P [Y i = ] = 1 p. Second description: Binomial

More information

First passage time for Brownian motion and piecewise linear boundaries

First passage time for Brownian motion and piecewise linear boundaries To appear in Methodology and Computing in Applied Probability, (2017) 19: 237-253. doi 10.1007/s11009-015-9475-2 First passage time for Brownian motion and piecewise linear boundaries Zhiyong Jin 1 and

More information

Computer Intensive Methods in Mathematical Statistics

Computer Intensive Methods in Mathematical Statistics Computer Intensive Methods in Mathematical Statistics Department of mathematics johawes@kth.se Lecture 16 Advanced topics in computational statistics 18 May 2017 Computer Intensive Methods (1) Plan of

More information

Multilevel Monte Carlo for Stochastic McKean-Vlasov Equations

Multilevel Monte Carlo for Stochastic McKean-Vlasov Equations Multilevel Monte Carlo for Stochastic McKean-Vlasov Equations Lukasz Szpruch School of Mathemtics University of Edinburgh joint work with Shuren Tan and Alvin Tse (Edinburgh) Lukasz Szpruch (University

More information

Solution of Stochastic Optimal Control Problems and Financial Applications

Solution of Stochastic Optimal Control Problems and Financial Applications Journal of Mathematical Extension Vol. 11, No. 4, (2017), 27-44 ISSN: 1735-8299 URL: http://www.ijmex.com Solution of Stochastic Optimal Control Problems and Financial Applications 2 Mat B. Kafash 1 Faculty

More information

Some Properties of NSFDEs

Some Properties of NSFDEs Chenggui Yuan (Swansea University) Some Properties of NSFDEs 1 / 41 Some Properties of NSFDEs Chenggui Yuan Swansea University Chenggui Yuan (Swansea University) Some Properties of NSFDEs 2 / 41 Outline

More information

Bisection Ideas in End-Point Conditioned Markov Process Simulation

Bisection Ideas in End-Point Conditioned Markov Process Simulation Bisection Ideas in End-Point Conditioned Markov Process Simulation Søren Asmussen and Asger Hobolth Department of Mathematical Sciences, Aarhus University Ny Munkegade, 8000 Aarhus C, Denmark {asmus,asger}@imf.au.dk

More information

Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio ( )

Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio ( ) Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio (2014-2015) Etienne Tanré - Olivier Faugeras INRIA - Team Tosca October 22nd, 2014 E. Tanré (INRIA - Team Tosca) Mathematical

More information

Computational statistics

Computational statistics Computational statistics Markov Chain Monte Carlo methods Thierry Denœux March 2017 Thierry Denœux Computational statistics March 2017 1 / 71 Contents of this chapter When a target density f can be evaluated

More information

The concentration of a drug in blood. Exponential decay. Different realizations. Exponential decay with noise. dc(t) dt.

The concentration of a drug in blood. Exponential decay. Different realizations. Exponential decay with noise. dc(t) dt. The concentration of a drug in blood Exponential decay C12 concentration 2 4 6 8 1 C12 concentration 2 4 6 8 1 dc(t) dt = µc(t) C(t) = C()e µt 2 4 6 8 1 12 time in minutes 2 4 6 8 1 12 time in minutes

More information

Introduction to Random Diffusions

Introduction to Random Diffusions Introduction to Random Diffusions The main reason to study random diffusions is that this class of processes combines two key features of modern probability theory. On the one hand they are semi-martingales

More information

On the quantiles of the Brownian motion and their hitting times.

On the quantiles of the Brownian motion and their hitting times. On the quantiles of the Brownian motion and their hitting times. Angelos Dassios London School of Economics May 23 Abstract The distribution of the α-quantile of a Brownian motion on an interval [, t]

More information

Hierarchical Bayesian Inversion

Hierarchical Bayesian Inversion Hierarchical Bayesian Inversion Andrew M Stuart Computing and Mathematical Sciences, Caltech cw/ S. Agapiou, J. Bardsley and O. Papaspiliopoulos SIAM/ASA JUQ 2(2014), pp. 511--544 cw/ M. Dunlop and M.

More information

The Smoluchowski-Kramers Approximation: What model describes a Brownian particle?

The Smoluchowski-Kramers Approximation: What model describes a Brownian particle? The Smoluchowski-Kramers Approximation: What model describes a Brownian particle? Scott Hottovy shottovy@math.arizona.edu University of Arizona Applied Mathematics October 7, 2011 Brown observes a particle

More information

Regular Variation and Extreme Events for Stochastic Processes

Regular Variation and Extreme Events for Stochastic Processes 1 Regular Variation and Extreme Events for Stochastic Processes FILIP LINDSKOG Royal Institute of Technology, Stockholm 2005 based on joint work with Henrik Hult www.math.kth.se/ lindskog 2 Extremes for

More information

Computational Complexity of Metropolis-Hastings Methods in High Dimensions

Computational Complexity of Metropolis-Hastings Methods in High Dimensions Computational Complexity of Metropolis-Hastings Methods in High Dimensions Alexandros Beskos and Andrew Stuart Abstract This article contains an overview of the literature concerning the computational

More information

Introduction to asymptotic techniques for stochastic systems with multiple time-scales

Introduction to asymptotic techniques for stochastic systems with multiple time-scales Introduction to asymptotic techniques for stochastic systems with multiple time-scales Eric Vanden-Eijnden Courant Institute Motivating examples Consider the ODE {Ẋ = Y 3 + sin(πt) + cos( 2πt) X() = x

More information

Cheng Soon Ong & Christian Walder. Canberra February June 2018

Cheng Soon Ong & Christian Walder. Canberra February June 2018 Cheng Soon Ong & Christian Walder Research Group and College of Engineering and Computer Science Canberra February June 2018 Outlines Overview Introduction Linear Algebra Probability Linear Regression

More information

17 : Markov Chain Monte Carlo

17 : Markov Chain Monte Carlo 10-708: Probabilistic Graphical Models, Spring 2015 17 : Markov Chain Monte Carlo Lecturer: Eric P. Xing Scribes: Heran Lin, Bin Deng, Yun Huang 1 Review of Monte Carlo Methods 1.1 Overview Monte Carlo

More information

Advances and Applications in Perfect Sampling

Advances and Applications in Perfect Sampling and Applications in Perfect Sampling Ph.D. Dissertation Defense Ulrike Schneider advisor: Jem Corcoran May 8, 2003 Department of Applied Mathematics University of Colorado Outline Introduction (1) MCMC

More information

The Azéma-Yor Embedding in Non-Singular Diffusions

The Azéma-Yor Embedding in Non-Singular Diffusions Stochastic Process. Appl. Vol. 96, No. 2, 2001, 305-312 Research Report No. 406, 1999, Dept. Theoret. Statist. Aarhus The Azéma-Yor Embedding in Non-Singular Diffusions J. L. Pedersen and G. Peskir Let

More information

Asymptotics for posterior hazards

Asymptotics for posterior hazards Asymptotics for posterior hazards Pierpaolo De Blasi University of Turin 10th August 2007, BNR Workshop, Isaac Newton Intitute, Cambridge, UK Joint work with Giovanni Peccati (Université Paris VI) and

More information

A concentration theorem for the equilibrium measure of Markov chains with nonnegative coarse Ricci curvature

A concentration theorem for the equilibrium measure of Markov chains with nonnegative coarse Ricci curvature A concentration theorem for the equilibrium measure of Markov chains with nonnegative coarse Ricci curvature arxiv:103.897v1 math.pr] 13 Mar 01 Laurent Veysseire Abstract In this article, we prove a concentration

More information

Lipschitz continuity for solutions of Hamilton-Jacobi equation with Ornstein-Uhlenbeck operator

Lipschitz continuity for solutions of Hamilton-Jacobi equation with Ornstein-Uhlenbeck operator Lipschitz continuity for solutions of Hamilton-Jacobi equation with Ornstein-Uhlenbeck operator Thi Tuyen Nguyen Ph.D student of University of Rennes 1 Joint work with: Prof. E. Chasseigne(University of

More information

Analysis of SPDEs Arising in Path Sampling Part I: The Gaussian Case

Analysis of SPDEs Arising in Path Sampling Part I: The Gaussian Case Analysis of SPDEs Arising in Path Sampling Part I: The Gaussian Case September 2, 2005 M. Hairer 1, A. M. Stuart 1, J. Voss 1,2, and P. Wiberg 1,3 1 Mathematics Institute, The University of Warwick, Coventry

More information

Turbulent Flows. g u

Turbulent Flows. g u .4 g u.3.2.1 t. 6 4 2 2 4 6 Figure 12.1: Effect of diffusion on PDF shape: solution to Eq. (12.29) for Dt =,.2,.2, 1. The dashed line is the Gaussian with the same mean () and variance (3) as the PDF at

More information

Markov Chain Monte Carlo Inference. Siamak Ravanbakhsh Winter 2018

Markov Chain Monte Carlo Inference. Siamak Ravanbakhsh Winter 2018 Graphical Models Markov Chain Monte Carlo Inference Siamak Ravanbakhsh Winter 2018 Learning objectives Markov chains the idea behind Markov Chain Monte Carlo (MCMC) two important examples: Gibbs sampling

More information

13 Notes on Markov Chain Monte Carlo

13 Notes on Markov Chain Monte Carlo 13 Notes on Markov Chain Monte Carlo Markov Chain Monte Carlo is a big, and currently very rapidly developing, subject in statistical computation. Many complex and multivariate types of random data, useful

More information

Higher order weak approximations of stochastic differential equations with and without jumps

Higher order weak approximations of stochastic differential equations with and without jumps Higher order weak approximations of stochastic differential equations with and without jumps Hideyuki TANAKA Graduate School of Science and Engineering, Ritsumeikan University Rough Path Analysis and Related

More information

AIMS Exercise Set # 1

AIMS Exercise Set # 1 AIMS Exercise Set #. Determine the form of the single precision floating point arithmetic used in the computers at AIMS. What is the largest number that can be accurately represented? What is the smallest

More information

Regularization by noise in infinite dimensions

Regularization by noise in infinite dimensions Regularization by noise in infinite dimensions Franco Flandoli, University of Pisa King s College 2017 Franco Flandoli, University of Pisa () Regularization by noise King s College 2017 1 / 33 Plan of

More information

Brief introduction to Markov Chain Monte Carlo

Brief introduction to Markov Chain Monte Carlo Brief introduction to Department of Probability and Mathematical Statistics seminar Stochastic modeling in economics and finance November 7, 2011 Brief introduction to Content 1 and motivation Classical

More information

Exercises. T 2T. e ita φ(t)dt.

Exercises. T 2T. e ita φ(t)dt. Exercises. Set #. Construct an example of a sequence of probability measures P n on R which converge weakly to a probability measure P but so that the first moments m,n = xdp n do not converge to m = xdp.

More information

Lecture 7 and 8: Markov Chain Monte Carlo

Lecture 7 and 8: Markov Chain Monte Carlo Lecture 7 and 8: Markov Chain Monte Carlo 4F13: Machine Learning Zoubin Ghahramani and Carl Edward Rasmussen Department of Engineering University of Cambridge http://mlg.eng.cam.ac.uk/teaching/4f13/ Ghahramani

More information

Short-time expansions for close-to-the-money options under a Lévy jump model with stochastic volatility

Short-time expansions for close-to-the-money options under a Lévy jump model with stochastic volatility Short-time expansions for close-to-the-money options under a Lévy jump model with stochastic volatility José Enrique Figueroa-López 1 1 Department of Statistics Purdue University Statistics, Jump Processes,

More information

Parameter estimation of diffusion models

Parameter estimation of diffusion models 129 Parameter estimation of diffusion models Miljenko Huzak Abstract. Parameter estimation problems of diffusion models are discussed. The problems of maximum likelihood estimation and model selections

More information

Web Appendix for Hierarchical Adaptive Regression Kernels for Regression with Functional Predictors by D. B. Woodard, C. Crainiceanu, and D.

Web Appendix for Hierarchical Adaptive Regression Kernels for Regression with Functional Predictors by D. B. Woodard, C. Crainiceanu, and D. Web Appendix for Hierarchical Adaptive Regression Kernels for Regression with Functional Predictors by D. B. Woodard, C. Crainiceanu, and D. Ruppert A. EMPIRICAL ESTIMATE OF THE KERNEL MIXTURE Here we

More information

Selected Exercises on Expectations and Some Probability Inequalities

Selected Exercises on Expectations and Some Probability Inequalities Selected Exercises on Expectations and Some Probability Inequalities # If E(X 2 ) = and E X a > 0, then P( X λa) ( λ) 2 a 2 for 0 < λ

More information

Gaussian Process Approximations of Stochastic Differential Equations

Gaussian Process Approximations of Stochastic Differential Equations Gaussian Process Approximations of Stochastic Differential Equations Cédric Archambeau Dan Cawford Manfred Opper John Shawe-Taylor May, 2006 1 Introduction Some of the most complex models routinely run

More information

Change detection problems in branching processes

Change detection problems in branching processes Change detection problems in branching processes Outline of Ph.D. thesis by Tamás T. Szabó Thesis advisor: Professor Gyula Pap Doctoral School of Mathematics and Computer Science Bolyai Institute, University

More information

Zig-Zag Monte Carlo. Delft University of Technology. Joris Bierkens February 7, 2017

Zig-Zag Monte Carlo. Delft University of Technology. Joris Bierkens February 7, 2017 Zig-Zag Monte Carlo Delft University of Technology Joris Bierkens February 7, 2017 Joris Bierkens (TU Delft) Zig-Zag Monte Carlo February 7, 2017 1 / 33 Acknowledgements Collaborators Andrew Duncan Paul

More information

Multimodal Nested Sampling

Multimodal Nested Sampling Multimodal Nested Sampling Farhan Feroz Astrophysics Group, Cavendish Lab, Cambridge Inverse Problems & Cosmology Most obvious example: standard CMB data analysis pipeline But many others: object detection,

More information

Lecture 12: Detailed balance and Eigenfunction methods

Lecture 12: Detailed balance and Eigenfunction methods Lecture 12: Detailed balance and Eigenfunction methods Readings Recommended: Pavliotis [2014] 4.5-4.7 (eigenfunction methods and reversibility), 4.2-4.4 (explicit examples of eigenfunction methods) Gardiner

More information

Numerical methods for solving stochastic differential equations

Numerical methods for solving stochastic differential equations Mathematical Communications 4(1999), 251-256 251 Numerical methods for solving stochastic differential equations Rózsa Horváth Bokor Abstract. This paper provides an introduction to stochastic calculus

More information

FE 5204 Stochastic Differential Equations

FE 5204 Stochastic Differential Equations Instructor: Jim Zhu e-mail:zhu@wmich.edu http://homepages.wmich.edu/ zhu/ January 20, 2009 Preliminaries for dealing with continuous random processes. Brownian motions. Our main reference for this lecture

More information

Model Specification Testing in Nonparametric and Semiparametric Time Series Econometrics. Jiti Gao

Model Specification Testing in Nonparametric and Semiparametric Time Series Econometrics. Jiti Gao Model Specification Testing in Nonparametric and Semiparametric Time Series Econometrics Jiti Gao Department of Statistics School of Mathematics and Statistics The University of Western Australia Crawley

More information

Elliptic Operators with Unbounded Coefficients

Elliptic Operators with Unbounded Coefficients Elliptic Operators with Unbounded Coefficients Federica Gregorio Universitá degli Studi di Salerno 8th June 2018 joint work with S.E. Boutiah, A. Rhandi, C. Tacelli Motivation Consider the Stochastic Differential

More information