Skorohod integration and stochastic calculus beyond the fractional Brownian scale (Q556245)

Stochastic integration is defined for a whole class of Gaussian integrators containing processes very close to fractional Brownian motion, as well as processes which may be much more irregular. More precisely, the integrators envisaged by the authors are centered Gaussian processes of the form \(B^{\gamma}(t)=\int_0^t\varepsilon(t-s)dW(s),\) where \(W\) is a standard Brownian motion on \(R_+\), whereas \(\varepsilon(r)=(d(\gamma^2)/dr)^{1/2}\) for a function \(\gamma\) such that \(\gamma^2\) is of class \(C^2\) everywhere on \(R_+\) except at \(0\) and \(d(\gamma^2)/dr\) is non-increasing. (Alternatively, the function \(\gamma\) may be defined only on some bounded interval \([0;T]\).) For \(H>0\) (arbitrarily small) and \(\gamma_H(r)=r^H\), one thus obtains a centered Gaussian process \(B^{\gamma_H}\) resembling fractional Brownian motion with parameter \(H\). Another important example of such integrators is provided by the logarithmic Brownian motion with parameter \(\beta\), defined by the authors to be the centered Gaussian process obtained when \(\gamma^2(r)=[\log(1/r)]^{-\beta},\varepsilon^2(r)=\frac{\beta}{r}[\log(1/r)]^{-\beta-1}.\) For such centered Gaussian integrators, the authors first define a Wiener integral (for non-random integrands). They then proceed to give a detailed study of the canonical Hilbert spaces associated with this Wiener integral, before using Malliavin calculus in order to define the corresponding Skorokhod integral as the adjoint operator of a derivative operator. Next, they establish an Itô formula for such Skorokhod integrals, and then the existence and some of the properties of local times, as well as the validity of a Tanaka formula. The final paragraph is a first study of some stochastic differential equations driven by centered Gaussian processes of the type mentioned above.