Malliavin Calculus

Malliavin Calculus

Malliavin calculus How far can you go with integration by parts? L. Gross L.Decreusefond Singular Greeks SDE’s Optimal Γ calculus (Condi- transport 2 tional) densities Dirichlet calculus Malliavin Stein- Antici- Malliavin- pative calculus Dirichlet calculus method Inde- pendent rand. var. Ind. increments Lévy Spatial Poisson Brownian fractional motion Brownian motion Martin- gales Sample-paths Itô calculus Rough paths SDE’s Stochastic integral 2 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Invariance yields IBP Invariance with respect to translation Z Z d f (x + τ)g(x + τ) dx = f (x)g(x) dx0 dτ R τ=0 R Integration by parts Z Z f 0(x)g(x) dx + f (x)g0(x) dx = 0 R R 3 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Brownian motion Definition The Brownian motion is the unique centered Gaussian process such that 1 E [B(t)B(s)] = min(t; s) = (t + s − jt − sj) 2 Independent and stationary increments Hölder (1/2 − ") sample-paths Finite quadratic variation 4 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Construction 1:2 0:0 0:4 1:0 0:2 0:1 − 0:8 0:0 0:2 − 0:6 0:2 − 0:3 0:4 − 0:4 − 0:6 0:4 0:2 − − 0:8 − 0:5 0:0 − 0:0 0:2 0:4 0:6 0:8 1:0 0:0 0:2 0:4 0:6 0:8 1:0 0:0 0:2 0:4 0:6 0:8 1:0 Donsker Karhunen-Loève Poisson [ ] p sin(π(n+ 1 )t) ( )−λ p1 P nt P 2 Nλ pt t j=1 Xj 2 Yn 1 n π(n+ 2 ) λ 5 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Standard Wiener measure Distribution of B supported by C0([0; 1]; R) Hol(1/2 − ") Wη;p for 0 < η − 1/p < 1/2 ZZ jf (t) − f (s)jp kf kp = ds dt Wη;p 1+ηp [0;1]2 jt − sj How to characterize a Gaussian measure on these Banach spaces ? 6 / 98 June 2018 Institut Mines-Télécom Malliavin calculus RKHS Definition (Reproducing Kernel Hilbert Space) n o H = span t ^ :; t 2 [0; 1] for the norm induced by the scalar product ht ^ :; s ^ :iH = t ^ s 7 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Identification Since Z s 1 1 2 t ^ s = 1[0;t](u) du = I0+ (1[0;t])(s) ) t ^ : 2 I0+ L ([0; 1]) 0 Z 1 and t ^ s = 1[0;t](u)1[0;s](u) du = h1[0;t]; 1[0;s]iL2 0 H as a Besov-Liouville space n o 1 2 H = span t ^ :; t 2 [0; 1] ' I0+ L ([0; 1]) with 1 _ _ kI0+ (h)kH = khkL2([0;1]) 8 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Abstract Wiener space Theorem For any choice of W among C; Hol(1/2 − ") or Wη;p, the triplet emb : H −! W is an AWS: H is dense in W , emb is radonifying and the standard Gaussian measure is characterized by 1 ∗ 2 E [exp(−hζ; !i ∗ )] = exp(− k emb ζk ) W ;W 2 H where ∗ emb emb W ∗ −−−!H∗ 'H −−! W 9 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Wiener integral ∗ emb emb W ∗ −−−!H∗ 'H −−! W Definition The Wiener integral is the isometric extension of the map ∗ ∗ 2 δ : emb (W ) ⊆ H −! L (P1/2) ∗ emb (ζ) 7−! hζ; !iW ∗;W : 10 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Identification ∗ ∗ emb ∗ emb W −−−!H 'H −−! W and H = I1;2 ∗ ht = emb "t must satisfy for ! 2 H ⊂ W and differentiable Z 1 ∗ _ !(t) = h"t; emb !iW ∗;W = hemb "t;!iH = ht(u)_!(u) du 0 _ ) ht(u) = 1[0;t](u) ) ht(u) = t ^ u Consequence ∗ ∗ 2 ∗ δ : emb (W ) ⊆ I1;2 −! L (P1/2) δ(t ^ :) = δ(emb (ζ)) = B(t) ∗ emb (ζ) 7−! hζ; !iW ∗;W 11 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Fractional Brownian motion Definition For any H in (0; 1), fBH (t); t ≥ 0g is the centered Gaussian process whose covariance kernel is given by VH 2 2 2 R (s; t) = E [B (s)B (t)] = s H + t H − jt − sj H H H H 2 where Γ(2 − 2H) cos(πH) V = : H πH(1 − 2H) stationary increments Hölder (H − ") sample-paths Finite 1/H-variation 12 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Some realizations H = 0:2 H = 0:8 0:10 2:0 1:5 0:05 1:0 0:00 0:5 0:05 − 0:0 0:10 − 0:5 − 0:15 − 1:0 − 0:20 − 1:5 − 0:0 0:2 0:4 0:6 0:8 1:0 0:0 0:2 0:4 0:6 0:8 1:0 H = 0:5 0:25 0:00 0:25 − 0:50 − 0:75 − 1:00 − 1:25 − 0:0 0:2 0:4 0:6 0:8 1:0 Figure: Sample-path example for H = 0:2, H = 0:5 (middle) and H = 0:8. 13 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Besov-Liouville spaces Definition (Riemann-Liouville fractional spaces) For α > 0, for f 2 Lp([0; 1]), Z t α 1 α−1 I0+ f (t) = (t − s) f (s) ds: Γ(α) 0 α p The space Iα,p is the set I0+ (L [0; 1]) equipped with the scalar product Z 1 α α p hI0+ f ; I0+ giIα,p = hf ; giL = f (s)g(s) ds: 0 14 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Sobolev embeddings For a > b > c and p; c − 1/p > 0 Wa;p ⊂ Ib;p ⊂ Wc;p ⊂ Hol(c − 1/p) 15 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Key lemma Square root of RH There exists KH such that Z 1 RH (t; s) = KH (t; r)KH (s; r) dr 0 2 6 7 5 6 1.5 4 5 4 1 3 3 2 0.5 2 1 0 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 K0:25(1;s) K0:5(1;s) K0:75(1;s) 16 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Properties Properties s K (t; s) = tH−1/2 K (1; ) H H t KH (t; s) = 0 if s > t K1/2(t; s) = 1[0;t](s) −|H−1/2j H−1/2 1 KH (t; s) = r (t − s) × C −function Warning ! Singularity of KH increases as H " 17 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Regularity p KH : L −! IH+1/2;p Z t f 7−! KH f (t) = KH (t; s)f (s) ds 0 is continuous. Essential remarks Z 1 RH (t; s) = KH (t; r)KH (s; r) dr ) RH (t; s) = KH KH (t;:) (s) 0 1 K1/2 = I0+ 18 / 98 June 2018 Institut Mines-Télécom Malliavin calculus RKHS Identification RKHS(BH ) = span fRH (t;:); t 2 [0; 1]g 2 = KH (L ) = IH+1/2;2 with scalar product Z 1 h _; _ i 2 = _( ) _ ( ) KH h KH g KH (L ) h s g s ds 0 19 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Identification Abstract Wiener space for fBm ∗ ∗ emb ∗ emb W −−−! IH+1/2;2 ' IH+1/2;2 −−! W _ For h = KH (h) 2 IH+1/2;2 ⊂ W ∗ h" ; i ∗ = h " ; i = ( ) t emb h W ;W emb t h IH+1/2;2 h t Z 1 = ( ; )_( ) = h ( ;:); _i 2 = h ( ( ;:)); i KH t s h s ds KH t h L KH KH t h IH+1/2;2 0 hence ∗ emb ("t) = KH (KH (t;:)) = RH (t;:) 20 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Consequences Definition The Wiener integral is the isometric extension of the map ∗ ∗ 2 δ : emb (W ) ⊆ I1;2 −! L (µW ) ∗ emb (ζ) 7−! hζ; !iW ∗;W Lemma δ(RH (t;:)) = BH (t) (δ(KH (1[0;t])); t 2 [0; 1]) is an ordinary BM 2 2 2 E δ(K (1 )) = kK (1 ))k = k1 k 2 = t H [0;t] H [0;t] IH+1/2;2 [0;t] L 21 / 98 June 2018 Institut Mines-Télécom Malliavin calculus In brief Embeddings and identifications ∗ ∗ emb ∗ ∗ W H = (IH+1/2;2) ' 2 KH emb L H = IH+1/2;2 W for W = C; or Wη;p with 0 < η − 1/p < H 1 ∗ 2 E [exp(−hζ; B i ∗; )] = exp − k emb ζk H W W 2 H 22 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Quasi invariance Theorem (Cameron-Martin) For any h 2 H, h i 1 E F B + emb(h) = E F(B ) exp δh − khk2 H H 2 H 23 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Girsanov theorem Theorem For u = KH u_ adapted, Z t Z t dQ 1 2 = exp u_ (s) dB(s) − u_ (s) ds P 2 d H Ft 0 0 then Z t ( ) − ( ; ) _ ( ) = ( ) LawQ BH t KH t s u s ds LawPH BH 0 24 / 98 June 2018 Institut Mines-Télécom Malliavin calculus Differentiation on W Obstacles Why not consider the Fréchet derivative 1 0 lim F(! + "! ) − F(!) ? "!0 " The Itô map is not continuous on W , so not Fréchet differentiable A random variable F is defined up to a negligeable set.

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    98 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