Identification of a noncausal Itô process from the stochastic Fourier coefficients (Q2637617)

The authors consider a Skorokhod-Itō process \[ X_t= X_0+ \int^t_0 a_s dW_s+ \int^t_0 b_s ds,\quad 0\leq t\leq 1, \] where \(W\) is a real Brownian motion and \(a_s(\omega)\) does not need to be adapted, but only to belong to the usual convenient Malliavin-Sobolev space. The purpose is here to recover the coefficients \(a_s(\omega)\) and \(b_s(\omega)\) by means of the only random Fourier coefficient \[ {\mathcal F}_k(X):= \int^1_0 \overline{e_k(t)}\,dX_t, \] where \(e_k(t):= \exp(2\pi\sqrt{-1} kt)\). Thus their result is the following satisfactory convergence in \(L^2\)-norm: for any \(n\in\mathbb{Z}\), \[ \int^1_0 a_t(\omega)\overline{e_n(t)}\,dt= \lim_{N\to\infty}\, {1\over 2N+1} \sum^N_{\ell=-N}{\mathcal F}_{n-\ell}(X)\times \int^1_0 \overline{e_\ell(t)}\,dW_t. \] This formula is inspired by the Fourier method introduced by \textit{P. Malliavin} and \textit{M. E. Mancino} [Finance Stoch. 6, No. 1, 49--61 (2002; Zbl 1008.62091)] and \textit{P. Malliavin} and \textit{A. Thalmaier} [Stochastic calculus of variations in mathematical finance. Berlin: Springer (2006; Zbl 1124.91035)]. The proof relies on a generalized integration by parts formula for Skorokhod integrals.