The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf...
Transcript of The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf...
![Page 1: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/1.jpg)
The Hopf algebraic structure of stochasticexpansions and efficient simulation
Kurusch Ebrahimi–Fard, Alexander Lundervold,Simon J.A. Malham, Hans Munthe–Kaas and Anke Wiese
York: 15th October 2012
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 1 / 32
![Page 2: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/2.jpg)
Introduction
SDE for yt ∈ RN in Stratonovich form
yt = y0 +
d∑i=0
∫ t
0Vi(yτ) dW i
τ.
Driven by a d-dimensional Wiener process (W 1, . . . ,W d );
Governed by drift V0 and diffusion vector fields V1, . . . ,Vd ;
Use convention W 0t ≡ t ;
Assume t ∈ R+ lies in the interval of existence.
Focus on solution series and strong simulation.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 2 / 32
![Page 3: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/3.jpg)
Accuracy: Problem
Background: Solutions only in exceptional cases given explicitly.Typically, approximations are based on the stochastic Taylorexpansion, truncated to include the necessary terms to achieve thedesired order (e.g. Euler scheme, Milstein scheme).
Question: Are there other series such that any truncation generates anapproximation that is always at least as accurate as the correspondingtruncated stochastic Taylor series, independent of the vector fields andto all orders?
We will call such approximations efficient integrators.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 3 / 32
![Page 4: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/4.jpg)
Flowmap
For the system of equations
dydt
= V (y)
we can define the flowmap ϕt : y0 7→ yt . Chain rule =⇒
ddt
f (y) = V (y) · ∇y f (y)︸ ︷︷ ︸V◦f◦y
⇔ddt
(f ◦ ϕt ◦ y0) = V ◦ f ◦ ϕt ◦ y0
⇔ddt
(f ◦ ϕt ) = V ◦ f ◦ ϕt
⇔ f ◦ ϕt = exp(tV ) ◦ f
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 4 / 32
![Page 5: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/5.jpg)
Flowmap for SDEs
Ito lemma for Stratonovich integrals =⇒
f ◦ yt = f ◦ y0 +
d∑i=0
∫ t
0Vi ◦ f ◦ yτ dW i
τ.
whereVi ◦ f = Vi · ∇y f .
Hence for flowmap we have
f ◦ ϕt = f ◦ id +
d∑i=0
∫ t
0Vi ◦ f ◦ ϕτ dW i
τ.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 5 / 32
![Page 6: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/6.jpg)
Flowmap series
Non-autonomous linear functional differential equation for ft B f ◦ ϕt
=⇒ set up the formal iterative procedure (with f (0)t = f ):
f (n+1)t = f +
d∑i=0
∫ t
0Vi ◦ f (n)
τ dW iτ,
Iterating =⇒
f ◦ϕt = f +
d∑i=0
(∫ t
0dW i
t
)︸ ︷︷ ︸
Ji (t)
Vi ◦ f +
d∑i ,j=0
(∫ t
0
∫ τ1
0dW i
τ2dW j
τ1
)︸ ︷︷ ︸
Jij (t)
Vi ◦ Vj︸ ︷︷ ︸Vij
◦f + · · · .
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 6 / 32
![Page 7: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/7.jpg)
Stochastic Taylor expansion
Stochastic Taylor expansion for the flowmap:
ϕt =∑
w∈A∗Jw (t) Vw .
w = a1 . . . an are words from alphabet A B {0,1, . . . ,d};
Scalar random variables
Jw (t) B∫ t
0· · ·
∫ τn−1
0dW a1
τn · · · dW anτ1
;
Partial differential operators Vw B Va1 ◦ · · · ◦ Van .
Encodes all stochastic and geometric information of system.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 7 / 32
![Page 8: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/8.jpg)
Basic Strategy
Is as follows:
1 Construct the series σt = f (ϕt );
2 Truncate the series σt to σt according to a grading;
3 Compute ϕt = f−1(σt ) =⇒ numerical scheme.
For example: Stochastic Taylor series implies
ϕt =∑
g(w)≤n
JwVw +∑
g(w)>n+1
JwVw
ϕt =∑
g(w)≤n
JwVw +∑
g(w)=n+1
E(Jw )Vw
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 8 / 32
![Page 9: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/9.jpg)
Example: Castell–Gaines method
Exponential Lie series ψt = logϕt given by
ψt = (ϕt − id) − 12(ϕt − id)2 + 1
3(ϕt − id)3 + · · ·
=
d∑i=0
JiVi +∑i>j
12(Jij − Jji)[Vi ,Vj ] + · · · .
Truncate, then across [tn, tn+1], we have
ψtn,tn+1 =
d∑i=0
(∆W i(tn, tn+1)
)Vi +
∑i>j
Aij(tn, tn+1) [Vi ,Vj ].
Thenytn+1 ≈ exp(ψtn,tn+1) ◦ ytn .
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 9 / 32
![Page 10: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/10.jpg)
Example: Castell–Gaines method recovery
For each path simulated ∆W i(tn, tn+1) and Aij(tn, tn+1) are fixedconstants =⇒ ψtn,tn+1 an autonomous vector field.
Thus, for τ ∈ [0,1] and with u(0) = ytn , solve
u′(τ) = ψtn,tn+1 ◦ u(τ).
Using a suitable high order ODE integrator generates u(1) ≈ ytn+1 .
Castell & Gaines: this method is asymptotically efficient in the sense ofN. Newton.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 10 / 32
![Page 11: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/11.jpg)
Hopf algebra representation
Stochastic Taylor series for the flowmap:
ϕ =∑w
w ⊗ w ,
lies in the product algebra (over K = R):
K〈A〉sh ⊗K〈A〉co.
See Reutenauer 1993.
(u ⊗ x)(v ⊗ y) = (u xxy v) ⊗ (xy),
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 11 / 32
![Page 12: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/12.jpg)
Hopf algebra of words
K〈A〉co:
Product, concatenation: u ⊗ v 7→ uv ;
Coproduct, deshuffle: w 7→∑
u xxy v=w u ⊗ v ;
Unit, counit and antipode S : a1 . . . an 7→ (−1)nan . . . a1.
K〈A〉sh:
Product, shuffle: u ⊗ v 7→ u xxy v ;
Coproduct, deconcatenation: w 7→ ∆(w) =∑
uv=w u ⊗ v ;
Unit, counit and antipode same.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 12 / 32
![Page 13: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/13.jpg)
Power series function
Suppose we apply function to the flow-map ϕ:
f (ϕ) =∑k>0
ck ϕk
=∑k>0
ck
∑w∈A∗
w ⊗ w
k
=∑k>0
ck
∑u1,...,uk∈A∗
(u1 xxy . . . xxy uk ) ⊗ (u1 . . . uk )
=∑
w∈A∗
∑k>0
ck
∑w=u1...uk
u1 xxy . . . xxy uk
⊗ w
=∑
w∈A∗F (w) ⊗ w
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 13 / 32
![Page 14: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/14.jpg)
Power series function recoding
In other words
f (ϕ) =∑
w∈A∗(F ◦ w) ⊗ w ,
where
F ◦ w =∑k>0
ck
∑u1,...,uk
w=u1...uk
u1 xxy . . . xxy uk .
=⇒ encode the action of f by F ∈ End(K〈A〉sh).
Note the stochastic Taylor series is F = id.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 14 / 32
![Page 15: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/15.jpg)
Shuffle convolution algebra
Embedding End(K〈A〉sh)→ K〈A〉sh ⊗K〈A〉co given by
X 7→∑w
X (w) ⊗ w ,
is an algebra homomorphism for the convolution product:
X ? Y = xxy ◦ (X ⊗ Y ) ◦∆.
We henceforth denoteH B End(K〈A〉sh).
Unit inH is ν and the antipode
S ? id = id ? S = ν.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 15 / 32
![Page 16: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/16.jpg)
Functions of the identity
F?(id) =∑k>0
ck id?k
log?(id) = J − 12J?2 + 1
3J?3− · · ·+
(−1)k+1
k J?k + · · ·
sinhlog?(id) = 12(id − S)
S = ν − J + J?2− J?3 + · · ·
sinhlog?(id) = J − 12J?2 + 1
2J?3− · · ·+ (−1)k+1 1
2J?k + · · · ,
where J B id − ν is the augmented ideal projector, it is the identity onnon-empty words and zero on the empty word:
J?k◦ w =
∑w=u1...uk
u1 xxy . . . xxy uk .
where all the words in the decomposition must be non-empty.KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 16 / 32
![Page 17: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/17.jpg)
Generating endomorphisms from endomorphisms
For any X ∈H:
f?(X ) =∑k>0
ck (X − ε ν)?k
log?(X ) =∑k>1
(−1)k+1
k(X − ν)?k
exp?(X ) =∑k>0
1k !
X?k
sinhlog?(X ) = 12
(X − X?(−1)
)coshlog?(X ) = 1
2
(X + X?(−1)
)
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 17 / 32
![Page 18: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/18.jpg)
Expectation endomorphism
Let D∗ ⊂ A∗ be the free monoid of words on D = {0,11,22, . . . ,dd}.
Definition (Expectation endomorphism)Linear map E ∈H such that
E : w 7→
tn(w)
2d(w)n(w)!11, w ∈ D∗,
0 11, w ∈ A∗\D∗.
Here d(w) is the number of non-zero consecutive pairs in w from D,n(w) = d(w) + z(w) where z(w) is the number of zeros in w .
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 18 / 32
![Page 19: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/19.jpg)
Inner product
Definition (Inner product)We define the inner product of X ,Y ∈H with respect to V to be
〈X ,Y 〉H B∑
u,v∈A∗E
(X (u) xxyY (v)
)(u, v).
The norm of an endomorphism X ∈H is ‖X‖H B 〈X ,X 〉1/2.
Motivation: if
xt =∑
w∈A∗X (w) Vw (y0) and yt =
∑w∈A∗
Y (w) Vw (y0),
our definition is based on the L2-inner product 〈xt , yt〉L2 = E (xt , yt ).
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 19 / 32
![Page 20: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/20.jpg)
Graded class
Definition (Grading map)This is the linear map g : K〈A〉sh → Z+ given by
g : w 7→ |w |.
Definition (Graded class)
For a given n ∈ Z+, we set Sn B{w ∈ K〈A〉sh : g(w) = n
}and
S6n B⊕k6n
Sk and S>n B⊕k>n
Sk .
A subset S ⊆ A∗ is of graded class if, for a given n ∈ Z+, S = Sn,S = S6n or S = S>n. We denote by πS : K〈A〉sh → S the canonicalprojection onto any graded class subspace S,
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 20 / 32
![Page 21: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/21.jpg)
Reversing Lemma
Lemma (M. & Wiese 2009)
For any pair u, v ∈ A∗, we have E (u xxy v) ≡ E((|S| ◦ u) xxy (|S| ◦ v)
).
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 21 / 32
![Page 22: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/22.jpg)
Lemma: properties
LemmaFor A = {0,1, . . . ,d} and graded class subspace S = Sn:
1 〈X ,Y 〉 = 〈|S| ◦ X , |S| ◦ Y 〉;
2 〈|S|, |S|〉 = 〈S,S〉 = 〈id, id〉;
3 〈sinhlog?(id), coshlog?(id)〉 = 0;
4 ‖id‖2 = ‖sinhlog?(id)‖2 + ‖coshlog?(id)‖2;
5 〈X ,E ◦ Y 〉 = 〈E ◦ X ,E ◦ Y 〉;
6 〈E ◦ id,E ◦ id〉 = 〈E ◦ |S|,E ◦ |S|〉 = 〈E ◦ S,E ◦ S〉;
7 〈E ◦ sinhlog?(id),E ◦ coshlog?(id)〉 = 0;
8 〈|S|, J?n〉 = 〈id, J?n
〉,
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 22 / 32
![Page 23: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/23.jpg)
Accuracy measurement
We measure the accuracy by ‖rt ◦ y0‖L2 where
rt B ϕt − ϕt .
A general stochastic integrator has the form
id B f?(−1)◦ πS6n ◦ f?(id).
The error is ‖R‖ in theH-norm where
R B id − id.
Definition (Pre-remainder)
Q B f?(id) − πS6n ◦ f?(id).
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 23 / 32
![Page 24: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/24.jpg)
Efficient integrator
Definition (Efficient integrator)We will say that a numerical approximation to the solution of an SDE isan efficient integrator if it generates a strong scheme that is moreaccurate in the root mean square sense than the correspondingstochastic Taylor scheme of the same strong order, independent of thegoverning vector fields and to all orders: i.e.
‖(id − E) ◦ R‖2 6 ‖(id − E) ◦ id‖2.
Sinhlog integrator of strong order n/2:
P B πS6n ◦ sinhlog?(id)
=(J − 1
2J?2 + · · ·+ 12(−1)n+1J?n
)◦ πS6n .
Q = sinhlog?(id) ◦ πS>n+1 .
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 24 / 32
![Page 25: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/25.jpg)
Sinhlog pre-remainder and remainder
h?(X ,+ν) =(X?2 + ν
)?(1/2)
= ν+ 12X?2
−18X?4 + · · ·
R = id − sinhlog?(−1)◦ P
= sinhlog?(−1)◦ (P + Q) − sinhlog?(−1)
◦ P= Q + h?(P + Q,+ν) − h?(P,+ν)
= Q + 12
((P + Q)?2
− P?2)
+ · · ·
= Q + 12
(P ?Q + Q ? P
)+ O
(Q?2
)= Q ◦ πSn+1 ,
since sinhlog?(id) = J + · · · : P ?Q =(J ◦ πS6n
)?
(J ◦ πS>n+1
).
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 25 / 32
![Page 26: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/26.jpg)
Sinhlog efficient
Reversing Lemma with S = Sn+1 =⇒
‖id‖2 = ‖Q‖2 + ‖coshlog?(id)‖2.
and ∥∥∥(id − E) ◦ id∥∥∥2
=∥∥∥(id − E) ◦Q
∥∥∥2+
∥∥∥(id − E) ◦ coshlog?(id)∥∥∥2.
Indeed, for any ε:
sinhlog?ε (id) = J + 12
∞∑k=2
(−1)k+1J?k + εJ?(n+1)
‖id‖2 = ‖Qε‖2 + ‖coshlog?(id)‖2 − ε
⟨id − S, J?(n+1)
⟩− ε2
∥∥∥J?(n+1)∥∥∥2.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 26 / 32
![Page 27: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/27.jpg)
Optimality Theorem
f?(X ; ε) B 12(X − εX?(−1)).
f?(−1)(X ; ε) = X + h?(X , ε ν),
TheoremFor every ε > 0 the class of integrators f?(id; ε) is efficent. Whenε = 1, the error of the integrator f?(id; 1) realizes its smallest possiblevalue compared to the error of the corresponding stochastic Taylorintegrator. Thus a strong stochastic integrator based on the sinhlogendomorphism is optimally efficient within this class, to all orders.
f?(id; ε) = 12(1 − ε)ν+ 1
2(1 + ε)J − 12ε
(J?2− J?3 + · · ·
).
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 27 / 32
![Page 28: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/28.jpg)
Optimality Theorem proof: step 1
h : (x , y) 7→ (x2 + y)1/2.
h(x + q, y) − h(x , y) =x
(x2 + y)1/2 · q + · · · ,
P B πS6n ◦ f?(id; ε)
R = id − id
= f?(−1)◦ (P + Q) − f?(−1)
◦ P= Q + h?(P + Q, ε ν) − h?(P, ε ν).
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 28 / 32
![Page 29: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/29.jpg)
Optimality Theorem proof: step 2
R = Q + h?(P + Q, ε ν) − h?(P, ε ν)
= Q +((P + Q)?2 + ε ν
)?(1/2)−
(P?2 + ε ν
)?(1/2)
= Q + 12(P?2 + ε ν)?(−1/2) ?
(P ?Q + Q ? P
)+ O
(Q?2
).
P = 12(1 − ε) ν+ O(J),
(P?2 + ε ν)?(−1/2) =(
12(1 + ε)
)−1ν+ O(J).
R = Q +1 − ε1 + ε
Q + O(J ?Q)
=2
1 + εQ ◦ πSn+1 .
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 29 / 32
![Page 30: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/30.jpg)
Optimality Theorem proof: step 3
On Sn+1:
‖id‖2 = 〈R + id − R,R + id − R〉
= ‖R‖2 + 2〈R, id − R〉+ 〈id − R, id − R〉
= ‖R‖2 + 2⟨id −
11 + ε
(id − εS),1
1 + ε(id − εS)
⟩+
⟨id −
11 + ε
(id − εS), id −1
1 + ε(id − εS)
⟩= ‖R‖2 + 1 −
1(1 + ε)2
(1 − 2ε〈id,S〉+ ε2
)= ‖R‖2 +
2ε(1 + ε)2
(1 + 〈id,S〉
).
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 30 / 32
![Page 31: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/31.jpg)
Optimality Theorem proof: step 4
R = Q + 12(P?2
− ν)?(−1/2) ?(P ?Q + Q ? P
)+ O
(Q?2
).
P = ν+ 12J?2 + O
(J?3
).
(P?2− ν)?(1/2) =
√
2 (P − ν)?(1/2) ?(ν+ 1
4(P − ν) + O((P − ν)?2
)).
R = Q +(J?(−1)
◦ πS6n
)?Q + O
(J ?Q
).
=⇒ order reduction.
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 31 / 32
![Page 32: The Hopf algebraic structure of stochastic expansions and ...simonm/talks/york.pdf · The Hopf algebraic structure of stochastic expansions and efficient simulation Kurusch Ebrahimi–Fard,](https://reader030.fdocuments.in/reader030/viewer/2022040916/5e8f72900fb8542a1a4da702/html5/thumbnails/32.jpg)
References
Ebrahimi-Fard, K., Lundervold, A., Malham, S.J.A., Munthe–Kaas,H. and Wiese, A. 2012 Algebraic structure of stochasticexpansions and efficient simulation, Proc. R. Soc. A.Lord, G., Malham, S.J.A. & Wiese, A. 2008 Efficient strongintegrators for linear stochastic systems, SIAM J. Numer. Anal.46(6), 2892–2919Malham, S.J.A., Wiese, A. 2008 Stochastic Lie group integrators,SIAM J. Sci. Comput. 30(2), 597–617Malham, S.J.A., Wiese, A. 2009 Stochastic expansions and Hopfalgebras. Proc. R. Soc. A 465, 3729–3749.Malham, S.J.A., Wiese, A. 2011 Levy area logistic expansion andsimulation, arXiv:1107.0151v1Newton, N.J. 1991, SIAM J. Appl. Math. 51, 542–567Reutenauer, C. 1993 Free Lie algebras, London MathematicalSociety Monographs New Series 7, Oxford Science Publications
KEF, AL, SM, HMK & AW () Hopf Algebras/SDEs York: 15th October 2012 32 / 32