DISCLAIMER

This note is based on Malliavin Calculus and Related Topics by David Nualart and is intended solely for personal academic study and should not be regarded as a substitute for the original text.

1 Wiener Chaos Decomposition

We seek a generalized form of Gaussian process, say changing the set of time indices T to a "more abstract" space H . Suppose H is a real separable Hilbert space with innrt product ⟨⋅,⋅⟩H and norm ‖h‖H for h∈H .

Finding Orthonormal Basis of Hn

Definition 1.1.8. (Generalized Hermite polynomials) We denote Λ be the set of all series a=(a1,a2,…),ai∈N whose indices are zero except finite number of them. Given multiindex a∈Λ , define a!:=∏i=1∞ai and |a|:=∑i=1∞ai . Generalized Hermite polynomial with multiindex a is defined as

Ha(x)=∏i=1∞Hai(xi)

Let {ei:i⩾1} be the orthonormal basis of infinite-dimensional Hilbert space H , we define a random variable using similar manner:

Φa:=a!∏i=1∞Hai(W(ei)),

which are the building blocks we are looking for. Moreover, Φa exhibits similar orthogonal properties as ordinary Hermite polynomials. Indeed, for any a,b∈Λ , we have

E[ΦaΦb]=E[a!b!∏i=1∞Hai(W(ei))Hbi(W(ei))]=a!b!∏i=1∞E[Hai(W(ei))Hbi(W(ei))]={1,a=b;0,a≠b.◻

To this point, we could probe that {Φa:|a|=n} is a complete orthonormal syetem in Hn :

Proposition 1.1.9. (Orthonormal system of Hn ) For any n⩾1 , the random variables {Φa:|a|=n} form a complete orthonormal system in Hn .

Proof. One could immediately see that {Φa:|a|=n} is a orthonormal system as E[ΦaΦb]=δab . On the other hand, consider a union of such sets of random variables instead:

⋃i=0n{Φa:|a|=i}.

Polynomials p(W(h1),…,W(hk)) in Pn could be approximated by W(e1),…,W(ej) , whose polynomials could then be expressed by random variables in {Φa,|a|⩽n} . However if we substitute n to n−1 , we have {Φa:|a|⩽n−1} forms a complete orthonormal system of Pn−1 therefore {Φa:|a|=n} forms a complete orthogonal system of Hn , as Pn=H0⊕⋯⊕Hn . Furthermore, {Φa:a∈Λ} is a complete orthonormal system of L2(Ω,F,P) . ◻